Email	fabi@vub.ac.be
Creation	Katrien Decq	21-Dec-1999
Last Update	Katrien Decq	12-Nov-2004

S. Gourvénec, G. Tomasi, C. Durville, E. Di Crescenzo, C.A. Saby, D.L. Massart, R. Bro & G. Oppenheim.

"CuBatch, a MATLAB interface for n-mode data analysis".

This paper describes the CuBatch interface software developed within the framework of the "NwayQual European Project". Originally built to deal with batch process data, CuBatch can be employed on the most disparate types of data. It includes different methods and algorithms for both multi-way and classical data analysis in different domains: curve resolution, exploratory analysis and regression. Moreover, it is intuitive as well as easy-to-use and results are provided as graphical outputs with numerous options. Since the list of methods is not exhaustive, CuBatch implements an open architecture allowing custom extensions. Some reference data sets suitable for learning and familiarizing with this interface and a detailed help in html format are also part of the software package. CuBatch may be seen as a general tool the use of which can be fruitful for both academic and industrial purposes.

E. Van Gyseghem, M. Jimidar, R. Sneyers, E. Verhoeven, D.L. Massart & Y. Vander Heyden.

"Selection of RP-HPLC columns with diverse selectivity towards the potential separation of impurities in drugs". J. Cromatogr. A

I. Stanimirova, M. Daszykowski, D.L. Massart, F. Questier, V. Simeonov & H. Puxbaum.

"Chemometrical exploration of the wet precipitation chemistry from the Austrian monitoring network (1988-1999)". J. Environm. Management

A. Detroyer, S. Stokbroekx, H. Bohets, W. Lorreyne, P. Timmerman, P. Verboven, D.L. Massart & Y. Vander Heyden.

" Fast monolithic micellar liquid chromatography, an alternative drug-permeability assessing method for high-throughput screening". Analytical chemistry

F. Questier, R. Put, D. Coomans, B. Walczak & D.L. Massart.

"The use of CART and multivariate regression trees for supervised and unsupervised feature selection" Chem. Intell. Lab Systems

X. Capron, B. Walczak, O. E. de Noord.

"Weighting of samples in multivariate regression model updating". Chem. Intell. Lab Systems.

D. Mangelings, I. Tanret, N. Matthijs, M. Maftouh, D.L. Massart & Y. Vander Heyden.

" Separation strategy for acidic chiral pharmaceuticals with capillary electrochromatography on polysaccharide stationary phases". Electrophoresis

T. Hancock, R. Put, D. Coomans, Y. Vander Heyden & Y. Everingham

"A Performance Comparison of Modern Statistical Techniques for Molecular Descriptor Selection and Retention Prediction in Chromatographic QSRR Studies" Chem. Intell. Lab. Systems 11/2004

S. Caetano, J. Aires-de-Sousa, M. Daszykowski and Y. Vander Heyden.

"Prediction of enantioselectivity using chirality codes and Classification and Regression Trees". Analytica Chimica Acta

   R. Put, Q.-S. Xu, D.L. Massart & Y. Vander Heyden.
"Multivariate Adaptive Regression Splines (MARS) in chromatographic quantitative structure-retention relationship studies". J. Chrom. A, 1055, 11-19 (2004)

   J. Luypaert, S. Heuerding, Y. Vander Heyden & D.L. Massart.
"The effect of preprocessing methods in reducing interfering variability from near-infrared measurements of creams". J. Pharm. Biom. Anal. , 36, 495-503 (2004)

N. Matthijs, S. Van Hemelryck, M. Maftouh, D.L. Massart & Y. Vander Heyden.

"Electrophoretic separation strategy for chiral pharmaceuticals using highly-sulfated and neutral cyclodextrins based dual selector systems". Anal. Chim. Acta, 525, 247-263 (2004)

A. Debunne, C. Vervaet, D. Mangelings & J.P. Remon.

"Compaction of enteric-coated pellets : influence of formulation and process parameters on tablet properties and in vivo evaluation". Eur. J. Pharm. Sciences, 22, 305-314 (2004)

The influence of formulation and compression parameters on the properties of tablets containing enteric-coated pellets, and on the integrity of the enteric polymer of the individual pellets after compression, was evaluated. From the D-optimality experimental design it could be concluded that the ratio of coated pellets to cushioning pellets (CoP/CuP) affected all tablet properties evaluated. Pellet size variation and altering the CoP/CuP ratio resulted in differences in in vitro tablet disintegration times.

Y. Vander Heyden, D. Mangelings, J. Van Brempt & H. Papen.

"Development and validation of an HPLC method with post-column derivatisation to assay N-acetylcysteine in plasma". Acta chromatographica, 14, 149-164 (2004)

An RPLC method has been developed enabling the determination of both low and high concentrations of N-acetylcysteine (NAC) in plasma. The compound is detected fluorimetrically after derivatisation. Validation of the method revealed that injection- and method repeatability were good. The linear range was adequate and the limit of quantification acceptable. Recovery of N-acetylcysteine from plasma samples was also acceptable. This method was applied to plasma samples from patients who had received very high doses of N-acetylcysteine.

S. Gourvénec & D.L. Massart.

" The orthogonal projection approach (OPA) and related methods in process monitoring". Anal. and Bioanal. Chem., 380, 373-375 (2004)

This communication is about the use of OPA for process monitoring. After a review of some classical methods of Multivariate Statistical Process Control (MSPC) accompagnied of many references, this short paper deals with the use of curve resolution methods for monitoring batch processes. Finally, description of OPA and example of industrial application of this method for this purpose are given.

J. Elek, D. Mangelings, F. Joó & Y Vander Heyden.

"Chemometric modelling of the catalytic hydrogenation of bicarbonate to formate in aqueous media". Reaction Kinetics and Catalysis LettersVol. 83 n°2, 321-328 (2004)

A central composite design was used to model the catalytic hydrogenation of aqueous bicarbonate. The results show good correlation between the measured and the modelled data. Response surface methodology was utilized to find the optimum for achieving the highest possible formic acid yield.

   F. Estienne, F. Despagne, B. Walczak, O.E. de Noord & D.L. Massart.
"A comparison of multivariate calibration techniques applied to experimental NIR data sets. Part III : robustness against instrumental perturbation conditions". Chemom. Intell. Lab. Systems

   I. Stanimirova, B. Walczak, D.L. Massart, V. Simeonov, C.A. Saby & E. di Crescenzo.
"STATIS, a 3-way method for data analysis. Application to environmental data". Chem. Intell. Lab. Systems

The present paper reports the results of a data exploration of 3-way environmental data with the use of STATIS ("Structuration des Tableaux A Trois Indices de la Statistique"). The method performance is compared with two commonly used methods in N-way data analysis, PARAFAC2 and Tucker3. STATIS appears as a very attractive 3-way exploratory method due to its robust properties, ability to deal with sets of 2-way tables that do not have the same number of columns or rows and lack of special requirements for data preprocessing.

D. Mangelings, M. Maftouh, Y. Vander Heyden & D.L. Massart.

"Enantioseparations by capillary electrochromatography: differences exhibited by normal- and reversed-phase versions of polysaccharide stationary phases". Electrophoresis, 25, 2808-2816 (2004)

The influence of using normal-phase and reversed-phase versions of four commercial polysaccharide stationary phases on chiral separations was investigated by means of capillary electrochromatography (CEC). Both versions of Chiralcel OD, OJ, and Chiralpak AD, AS were tested for the separation several compounds. Different background electrolytes were used, two at low pH and three at high pH. This set-up allowed evaluating differences between both stationary-phase versions and between mobile-phase compositions on a chiral separation. Duplicate CEC columns of each stationary phase were in-house prepared and tested, giving information about the inter-column reproducibility. The reversed-phase versions of the current commercial polysaccharide stationary phases are found to be best for reversed-phase applications in CEC, even though at high pH no significant differences were seen between both versions. For acidic compounds, it was seen that an ammonium formate electrolyte performed best. For basic, bifunctional and neutral compounds, no significant differences between the three tested electrolytes were observed at high pH. Therefore, a phosphate buffer is preferred as electrolyte because of its buffering capacities.

   S. Gourvénec, X. Capron & D.L. Massart.
"Genetic algorithms (GA) applied to the orthogonal projection approach (OPA) for variable selection". Anal. Chim. Acta, 519, 11-21 (2004)

Multivariate curve resolution (MCR) and especially the Orthogonal Projection Approach (OPA) can be applied to NIR spectroscopic data and were proved to be suitable for batch process monitoring. To improve the quality of the on-line monitoring, it is interesting to get as many as possible spectra in a given period of time. Nevertheless, due to some hardware limitations, it is not possible to acquire more than a certain number of spectra in this given period of time. Wavelength selection could be a good way to deal with this problem since it decreases size, and consequently the acquisition time, of each recorded spectrum. This paper details an application of genetic algorithms (GA) coupled with OPA for such purpose.

A. Nguyen Minh Nguyet, A.M. van Nederkassel, L. Tallieu, S. Kuttatharmmakul, E. Hund, Y. Hu, J. Smeyers-Verbeke & Y. Vander Heyden.

"Statistical method comparison: short- and long-column HPLC assays of ketoconazole and formaldehyde in shampoo". Anal. Chim. Acta, 516, 87-106 (2004)

N. Matthijs, C. Perrin, M. Maftouh, D.L. Massart & Y. Vander Heyden.

"Definition and system implementation of strategies for method development of chiral separations in normal- or reversed-phase liquid chromatography using polysaccharide-based stationary phases". J. Chromat. A, 1041, 119-133 (2004)

This paper proposes strategies in normal- and reversed-phase liquid chromatography (NP-HPLC or NPLC and RP-HPLC or RPLC), developed using three polysaccharide-based stationary phases (Chiralcel OD, Chiralpak AD en Chiralcel OJ). Those strategies are implemented in a knowledge-based system for the chiral separation of drug enantiomers. Each strategy includes a screening and an optimisation stage using experimental design. The screening stage allows a fast evaluation of separation possibilities and enantioselectivity for many drugs in a short period of time, while the optimisation stage gives the opportunity to enhance, if needed, the initially obtained separation. Different examples demonstrate the effectiveness of the strategies for fast method development.

D. Mangelings, M. Maftouh, D.L. Massart & Y. Vander Heyden.

"Comparison of the miniaturized techniques capillary electrochromatography and capillary liquid chromatography for the chiral separation of chlorthalidone" J. Pharm. & Biom. Analysis, 35, 789-799 (2004)

The aim of this study was to compare the techniques capillary electrochromatography (CEC) and capillary liquid chromatography (CLC) for the chiral separation of chlorthalidone. To perform the chiral separation of this compound, hydroxypropyl-b-cyclodextrin was used as chiral selector in the mobile phase, while an achiral stationary phase was used. Earlier, this separation was already optimised in CEC. Now, the separation was optimised in CLC. By means of a central composite design the influence of the organic modifier content and the cyclodextrin concentration on the separation were studied . Optimal separation conditions were determined from the response surface contour plots. When these conditions were compared with the CEC optimisation, we can see that less chiral selector was used, faster separations were obtained and better repeatability was observed in comparison with its electrical-driven counterpart.

   Q.-S. Xu, M. Daszykowski, B. Walczak, F. Daeyaert, M.R. de Jonge, J. Heeres, L.M.H. Koymans, P.J. Lewi, H.M. Vinkers, P.A. Janssen & D.L. Massart.
"Multivariate adaptive regression splines - studies of HIV reverse transcriptase inhibitors". Chem. Intell. Lab. Systems, 72, 27-34 (2004)

The application of Multivariate Adaptive Regression Splines (MARS) for modelling the biological activity of Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTIs) is introduced. The data consist of the biological activities and the computed interaction energies with the amino acids forming the NNRTI binding pocket. The MARS model obtained yields agreeable prediction performance and is explained in detail.

   I. Stanimirova, B. Walczak, D.L. Massart & V. Simeonov.
"A comparison between two robust PCA algorithms". Chem. Intell. Lab. Systems, 71, 83-95 (2004)

   Q.S. Xu, S. de Jong, P. Lewi & D.L. Massart.
"Partial least squares regression with Curds and Whey". Chemom. Intell. Lab. Systems, 71, 21-31 (2004)

To make full use of information between correlated responses in calibration, a new method called PLS-C&W is developed. It combines PLS regression with a canonical analysis between the PLS estimates and the responses. The estimates are suitably compressed in the canonical co-ordinate system using generalized cross validation. The results obtained from five data sets show that PLS-C&W methods yield calibration models of better prediction performance.

A. Detroyer, Y. Vander Heyden, K. Reynaert & D.L. Massart.

"Evaluating 'fast' micellar monolithic liquid chromatography for high-throughput QSRR screening". Anal. Chem., 76, 1903-1908 (2004)

Monolithic silica columns were tested in micellar liquid chromatography. Micellar methods are utilized in high-throughput quantitative structure-retention relationships to estimate the octanol-water partition coefficient, log P. The column's ability to function at higher flow rates might speed up these chromatographic methods. Therefore, the elution of basic pharmaceutical substances was determined on a particle-based and a monolithic column, both with and without micellar medium in the mobile phase. The extent to which these methods differ was examined in the context of high-throughput determination of log P. Combining monolithic columns with micellar media leads to faster log P, and possibly even better permeability, predictions.

   M. Daszykowski, B. Walczak, Q.-S. Xu, F. Daeyaert, M.R. de Jonge, J. Heeres, L.M.H. Koymans, P.J. Lewi, H.M. Vinkers, P.A. Janssen & D.L. Massart.
"Classification and regression trees - studies of HIV reverse transcriptase inhibitors". J. Chem. Inf. Comp. Sci., 44, 716-726 (2004)

The application of Classification And Regression Trees (CART) is presented for the analysis of biological activity of Non-Nucleoside Reverse Transcriptase Inhibitors (NNRTIs). The data consist of the biological activities, expressed as pIC50, of 208 NNRTIs against wild-type HIV virus (HIV-1) and four mutant strains (181C, 103N, 100I, 188L) and the computed interaction energies with the Reverse Transcriptase (RT) binding pocket. CART allows to explain the observed biological activity of NNRTIs in terms of interactions with individual amino acids in the RT binding pocket, i.e., the original data variables.

D. Mangelings, C. Perrin, D.L. Massart, M. Maftouh, S. Eeltink, W.Th. Kok, P.J. Schoenmakers & Y. Vander Heyden.

"Optimisation of the chlorthalidone chiral separation by capillary electrochromatography using an achiral stationary phase and cyclodextrin in the mobile phase". Anal. Chim. Acta, 509, 11-19 (2004).

The separation of chlorthalidone enantiomers was optimised in capillary electrochromatography on an achiral stationary phase when adding a chiral selector, hydroxypropyl-b-cyclodextrin, to the mobile phase. The goal was to investigate the feasibility of modeling retention times and resolution when regular replacement of columns is required due to breakage. The influence of the chiral selector concentration and the organic modifier content on the responses was modeled. A second-order polynomial model was built and optimal separations conditions were predicted using Derringer's desirability function. The optimum was found at 33 mM cyclodextrin and 16 % (v/v) acetonitrile. Measured and predicted responses were comparable.

P. Dehouck, D. Visky, Y. Vander Heyden, E. Adams, Z. Kovács, B. Noszál, D.L. Massart & J. Hoogmartens.

"Characterisation of reversed-phase liquid-chromatographic columns by chromatographic tests. Comparing column classification based on chromatographic parameters and column performance for the separation of acetylsalicylic acid and related compounds". J. Chromat. A, 1025, 189-200 (2004)

Selection of RP-LC columns with suitable selectivity for a given analysis is difficult. Official compendia for drug analysis only give a general description of the stationary phase in the operating procedure of a method. The need for a general test method to characterise RP-LC columns has been rising since the 1970s. A project to define a chromatographic procedure characterising RP-LC columns was started earlier. A procedure to measure test parameters was introduced and a classification of the columns was obtained. This paper focuses on correlating the column classification with the selectivity obtained for a real separation. The separation of acetylsalicylic acid (aspirin) and related compounds was performed according to the Ph. Eur. monograph. It was examined whether the classes of columns, determined using test parameter results contain either suitable or unsuitable supports for the aspirin separation. The system suiability test of the Ph. Eur. to distinguish between suitable or unsuitable columns was also evaluated.

   R. Conde, J. Smeyers-Verbeke & D.L. Massart.
"Iterative least squares method for regression analysis with censored data". TrAC, 23, N° 2, 146-156 (2004)

The iterative least squares method (ILSM) is studied for regression analysis with censored data. Simulated data were used to test the performance of this approach, which was compared with the classical least squares (LS) method applied to data in which the censored observations are ignored.The simple ILSM algorithm can make the statistical analysis involving censored data more likely to be used by analysts, as is shown on real data sets from chromatography and pharmaceutical technology.

   T. Iványi, K. Pál, I. Lázár, D.L. Massart & Y. Vander Heyden.
"Application of tetraoxadiaza-crown ether derivatives as chiral selector modifiers in capillary electrophoresis". J. Chromatogr. A, 1028, 325-332 (2004)

Two new diaza-crown ether derivatives have been synthesized and tested as potential chiral selectors in capillary electrophoresis (CE) for the chiral separation of five amino acid derivatives. The individual use of the selectors did not lead to chiral differentiation. However, they enhanced the enantioselective effect of different cyclodextrins in dual selector systems. The effect of different substituted diaza-crown ether derivatives on the separation results obtained in dual systems with cyclodextrins was reported.

   E. Vasbinder, G. Van der Weken, Y. Vander Heyden, W.R.G. Baeyens, A. Debunne, J.P. Remon & A.M. García-Campańa.
"Quantitative determination of p-aminosalicylic acid and its degradation product m-aminophenol in pellets by ion-pair high performance liquid chromatography applying the monolithic Chromolith Speedrod RP-18e column". Biomed.Chromatogr., 18, 55-63 (2004).

An ion-pair HPLC method was developed for the simultaneous determination of p-aminosalicylic acid (PAS) and m-aminophenol (MAP) in a newly developed multiparticular drug delivery system. The separation was performed on a new packing material consisting of monolithic rods of highly porous silica. The method was validated and proved to be useful for stability testing of a new dosage form. Separation efficicncy was comparcd between the new paking material and a conventional rcversed-phasc cartridge. A robustncss test was carried out on both columns and different separation parameters (retention, resolution, run time, temperature) were determined.

E. Van Gyseghem, I. Crosiers, S. Gourvénec, D.L. Massart & Y. Vander Heyden.

"Determining orthogonal and similar chromatographic systems from the injection of mixtures in liquid chromatography-diode array detection and the interpretation of correlation coefficients color maps". J. Chromatogr. A, 1026, 117-128 (2004).

Generically applicable orthogonal chromatographic systems might be helpful tools as potential starting points in method development to separate impurities and the active substance in drugs with unknown impurity profiles. The orthogonality of 38 chromatographic systems was evaluated from weighted-average-linkage dendrograms and color maps that are both based on the correlation coefficients between the retention factors on the different systems. On each chromatographic system, 68 drug substances were injected as mixtures of 3 or 4 components to increase the throughput. The (overlapping) peaks were identified and resolved with the peak purity algorithm, orthogonal projection approach (OPA). The visualization techniques applied allowed a simple evaluation of orthogonal and (groups of) similar systems.

M.H. Zhang, Q.S. Xu, & D.L. Massart.

"Averaged and weighted average partial least squares". Anal. Chim. Acta, 504, 279-289 (2004)

Two alternative PLS methods, averaged PLS and weighted average PLS, are compared with the classical PLS. The prediction abilities of the alternative methods are comparable to that of the classical PLS but they do not require to determine how many components should be included in the model. They are also more robust since the quality of prediction depends less on a good choice of the number of components to be included. In addition, weighted average PLS is also compared with the weighted average part of LOCAL, a published method that also applies weighted average PLS, with however an entirely different weighting scheme.

   M. H. Zhang, J. Luypaert, J.A. Fernández Pierna, Q.S. Xu and D. L. Massart
"Determination of total antioxidant capacity in green tea by near-infrared spectroscopy and multivariate calibration". Talanta, 62-1, 25-35 (2004)

A principal component regression model to predict the total antioxidant capacity in green tea using near-infrared (NIR) spectroscopy is built. Different outlier detection methods are compared. The predictability of the final model is comparable to the precision of the reference method.

P. Dehouck, E. Van Looy, E. Haghedooren, K. Deckers, Y. Vander Heyden, E. Adams, E. Roets and J. Hoogmartens

"Analysis of erythromycin and benzoylperoxide in topical gels by liquid chromatography"., J. Chrom. B, 794, 293-302 (2003)

Gels containing erythromycin and benzoylperoxide are used to treat acne vulgaris. A method was developed to determine both erythromycin and benzoylperoxide. Erythromycin was extracted from the gel in conditions where the oxidative power of benzoylperoxide was neutralised by addition of ascorbic acid and this extract was analysed. A second extraction procedure was developed for benzoylperoxide. The selectivity, repeatability, linearity and recovery of both methods were examined. Special attention was given to the recovery and the uncertainty on the recovery. This allowed evaluation of the bias of the extraction. The method was used to examine the stability of a gel for topical use.

C. Perrin, H. Fabre, D.L. Massart & Y. Vander Heyden.

"Influence of peak measurement parameters on the quality of chiral electrophoretic separations". Electrophoresis, 24, 2469-2480 (2003).

The effect of nine peak measurement/analysis parameters on chiral capillary electrophoresis (CE) was studied. The parameters pertained UVdetection (detection wavelength, reference wavelength, and wavelength bandwidths), signal processing (data acquisition rate, type and amount of filtering) and peak detection (detection threshold and peak width). The influence of these factors on the chiral separation of enantiomers was studied at two different concentrations (i.e., at high and low signal-to-noise (S/N) ratio). The characteristics considered were the resolution between the two enantiomers, the peak areas, and the S/N ratio. A D-optimal design was used as screening design to identify the most critical parameters. Afterwards, a modelling of different responses as a function of these critical parameters was performed based on the results of a face-centered central composite design. The signal-processing parameters should be carefully selected when developing a CE separation since very important variations in the separation, the S/N ratio and the peak area of the substances can occur by setting these parameters at different levels. The detection wavelength should also be carefully chosen for optimal peak area measurement. These parameters become more important with decreasing concentration of the analytes (i.e., low S/N ratio). This study showed that the peak measurement/analysis parameters should be optimized as the chemical and physical parameters of a method. They also should be well specified to allow a good transfer of a method from one instrument to another.

   J.O. De Beer, P. Baten, C. Nsengyuma & J. Smeyers-Verbeke.
"Measurement uncertainty from validation and duplicate analysis results in HPLC analysis of multivitamin preparations and nutrients with different galenic forms". J. Pharm. & Biomed. Analysis, 32, 767-811 (2003)

An approach to calculate the measurement uncertainty in the HPLC analysis of several vitamins in multivitamin preparations is described. It is examined if duplicate analysis results, on different lots of an effervescent tablet, spread over several points of time, might contribute to calculate the measurement uncertainty of the HPLC method used and if the established uncertainty is acceptable in the assessment of compliance with the legal content limits. Analysis of variance (ANOVA) and precision calculations, are applied to estimate precision components, necessary to derive the measurement uncertainty. In the second instance it is demonstrated to which extent the fully validated method is applicable to other galenic forms and which specific modifications are involved. By means of duplicate analysis results, acquired from a large series of real samples over a considerable period of time and classified according to their similarity in content, galenic forms and matrices, measurement uncertainty calculations are shown.

C. Holvoet, J. Plaizier-Vercammen, Y. Vander Heyden, M. Gabriëls & F. Camu

"Preparation and in-vitro release rate of fentanyl-cyclodextrin complexes for prolonged action in epidural analgesia". Int. J. Pharmaceutics, 265, 13-26 (2003)

Fentanyl was complexed with cyclodextrin derivatives to obtain parenteral solutions able to provide prolonged analgesia. Three cyclodextrins (CDs) were used : hydroxypropyl-β-cyclodextrin (HP-β-CD), sulfobutylether-β-cyclodextrin (SBE-7-β-CD), and maltosyl-β-cyclodextrin (malt-β-CD). The inclusion capacity of HP-β-CD was determined from phase-solubility diagrams at pH 6.5, 7.2 and 8.0, and those of SBE-7-β-CD and of malt-β-CD at pH 8.0. Solubility of fentanyl increased linearly (i) as a function of the CD concentration, and (ii) with decreasing pH. Complexation was highest with HP-β-CD and malt-β-CD. The CD concentration was calculated to obtain a fentanyl-CD formulation, with the desired amount of free fentanyl as loading dose in solution and the rest complexed with CD, as reservoir for prolonged action. Release rate was evaluated in static and dynamic conditions. For both modes the release rate of fentanyl decreased as a function of CD concentration, due to complex formation of fentanyl, which suggests the possibility to provide prolonged pharmacodynamic effects in vivo.

W.R.G. Bayens, G. Van der Weken, Y. Vander Heyden, V. Van Bossche, H.Y. Aboul-Enein, A.M. García-Campańa & E. Smet.

"Investigation of the chiral separation of cisapride on chiralcel OJ and OJ-R columns". Anal. Chim. Acta, 498, 9-24 (2003).

The Chiralcel OJ and OJ-R columns were tested for their enantiodiscriminating properties towards cisapride. The effect of mobile phase composition upon resolution and analysis time was investigated on both phases by means of an experimental design. Definition of useful combinations of variables resulting in acceptable enantioselective separation by simultaneously optimising analysis time was aimed at. The methods obtained were validated in terms of linearity, detection limit and precision. Both systems were applied to the quantitative determination of cisapride in pharmaceutical preparations.

   J.A. Fernández Pierna, L. Jin, M. Daszykowski, F. Wahl & D.L. Massart
"A methodology to detect outliers/inliers in prediction with PLS". Chem. Intel. Lab. Syst., 68, 17-28 (2003)

The homogeneity of data should be studied to guarantee the detection of outliers and inliers in prediction with a PLS model. An automatic methodology in three steps, with the possibility for visual checking, to detect these objects was developed. First the objects are mapped from an n-dimensional to a 2-dimensional space using Sammon's mapping. Then, clusters in the calibration space are detected using a density-based method, and finally, the convex hull method is applied to each cluster to detect outliers/inliers in new samples. Several case studies were carried out. The results show that these three steps make the detection of outliers and inliers for prediction easier and more accurate than classical methods.

   S. Gourvénec, J.A. Fernández Pierna, D.L. Massart & D.N. Rutledge.
"An evaluation of the PoLiSh smoothed regression and the Monte Carlo Cross-Validation for the determination of the complexity of a PLS model". Chem. Intel. Lab. Syst., 68, 41-51 (2003).

A crucial point of the PLS algorithm is the selection of the number of factors to avoid overfitting. The leave-one-out cross-validation is usually used, but in practice, it is found that often too many components are retained. In this study, the Monte Carlo Cross-Validation (MCCV) and the PoLiSh smoothed regression are used and compared with the adjusted Wold's R criterion.

   M.H. Zhang, Q.S. Xu & D.L. Massart
"Robust principal components regression based on principal sensitivity vectors". Chem. Intel. Lab. Syst., 67, 175 - 185 (2003)

A robust method called robust principal components regression based on principal sensitivity vectors (RPPSV) is developed for outlier detection in regression. The method is evaluated by its outlier detection ability and the root mean square error of prediction (RMSEP) for a test set of simulated data based on a real results. The RPPSV is compared with several robust outlier diagnostic methods. It shows that when the data set is lowly contaminated, the RPPSV has good outlier detection ability, especially for bad leverage points, and its RMSEP value is comparable to the other methods. When the data set is highly contaminated, the RPPSV has the best outlier detection ability and the lowest RMSEP.

P. Dehouck, Y. Vander Heyden, J. Smeyers-Verbeke, D.L. Massart, R.D. Marini, P. Chiap, Ph. Hubert, J. Crommen, W. Van de Wauw, J. De Beer, R. Cox, G. Mathieu, J.C. Reepmeyer, B. Voigt, O. Estevenon, A. Nicolas, A. Van Schepdael, E. Adams & J. Hoogmartens

"Interlaboratory study of a liquid chromatography method for erythromycin: determination of uncertainty". J. Chromatography A, 1010, 63 - 74 (2003)

A method for the analysis of erythromycin by liquid chromatography has previously been developed. After validation in one laboratory, the method was now validated in an interlaboratory study. The data derived in the interlaboratory study can be used to make an uncertainty statement as well. Eight laboratories participated in this interlaboratory study. The set-up allowed the determination of the repeatability variance and the between-laboratory variance. Combination of both results in the reproducibility variance. It has been shown how these data can be used in future by a single laboratory that wants to make an uncertainty statement concerning the same analysis.

D. Mangelings, N. Hardies, M. Maftouh, C. Suteu, D.L. Massart & Y. Vander Heyden

"Enantioseparations of basic and bifunctional pharmaceuticals by capillary electrochromatography using polysaccharide stationary phases". Electrophoresis, 24, 2567-2576 (2003)

A fast screening strategy was developed in capillary electrochromatography (CEC) for the chiral separation of basic and bifunctional compounds. The screening conditions were determined on polysaccharide chiral stationary phases using 15 pharmaceutical compounds. A generic mobile phase for basic and bifunctional compounds was determined. The testing of 20 additional compounds showed that the mobile phase allowed enantioselectivity for 86% of the compounds. A comparison of CEC and reversed-phase liquid chromatography results was attempted to demonstrate the CEC potential for chiral method development.

B. Rambali, G. Verreck, L. Baert. and D.L. Massart

"Itraconazole formulation studies of the melt-extrusion process with mixture design". Drug dev. & ind. pharm., 29 (6), 641-652 (2003)

Itraconazole is poorly water soluble. A method to increase the solubility is formation of a solid dispersion. Our purpose was to develop a 40% w/w itraconazole formulation through solid dispersion formation, using hydroxypropyl-β-cyclodextrin (HP-β-CD) and hydroxypropylmethylcellulose (HPMC) as mixture components. The dispersion was obtained by melt-extrusion. A D-optimal mixture design was applied to develop the itraconazole formulation. The itraconazole fractions varied between 20% w/w and 50% w/w in the design and the HPMC and HP-β-CD fractions between 10% w/w and 60% w/w. The formulation was optimized by producing clear extrudates, minimizing the torque, maximizing the glass transition temperature and the apparent itraconazole solubility in 0,1 N HCl. Regression models were developed for the torque, glass transition temperature and the apparent itraconazole solubility of itraconazole. The optimal itraconazole formulation consisted of 45% w/w HPMC and 15% HP-β-CD w/w.

C. Perrin, H. Fabre, M. Maftouh, D. L. Massart & Y. Vander Heyden

"Robustness testing of chiral separations by capillary electrophoresis using highly-sulfated cyclodextrins". J. Chromatography A, 1007, 165 - 177 (2003)

The robustness of a generic method for chiral separation in capillary electrophoresis using highly-sulfated cyclodextrins and the "short-end injection technique" was studied. We focused on the robustness of the separations and not of the quantitative analysis of the enantiomers. The robustness was evaluated for the enantiomeric separation of a basic , a neutral and an acidic compound. The influence of eight factors was studied using a 11-factor, 12-experiment Plackett-Burman design. Statistical interpretation of the factor effects on selectivity and resolution was performed. The separations of the three compounds could be considered as rather robust as the factor effects were generally not significant (α = 0.05).

   L. Jin, J.A. Fernández Pierna, Q. Xu, F. Wahl, O.E. de Noord, C.A. Saby & D.L. Massart
"Delaunay triangulation method for multivariate calibration". Anal. Chim. Acta, 488, 1-14 (2003)

The Delaunay triangulation (DT) method is proposed as a new local multivariate calibration method. The study compares the performance of the DT method with the global methods, principal component regression and partial least squares and the local ones, locally weighted regression and the law of mixtures . The DT method gives similar results when the root mean square error for prediction value is compared. However, the DT method requires fewer components than PCR and PLS.

Y. Vander Heyden.

Robuustheidstesten voor biomedische en farmaceutische analysemethoden (robustness tests for biomedical and pharamceutical analysis methods), Verhandelingen van de Koninklijke Academie voor Geneeskunde in België, LXV, 1, 47-73 (2003) (in Dutch).

In biomedical and pharmaceutical analysis, especially in the pharmaceutical industry, the quality of the results is strictly evaluated and controlled. Newly developed methods of analysis are subjected to a strict and vast method validation. One of the validation criteria is the determination of the robustness of the method. This is done by means of a robustness test. In this manuscript the different steps to set up and interprete a robustness test are discussed and illustrated with a number of examples.

B. Rambali, S. Van Aelst, L. Baert & D.L. Massart.

"Using deepest regression method on the granulation process". Int. J. Pharmaceutics, 258, 85-94 (2003).

The deepest regression method was applied to estimate the granule size of unsuccessful fluidized bed granulation runs. Data from a previous study on optimization of fluidized granulation process, wherein 8 of the 30 runs did not succeeded due to overwetting of the powder bed were used. The "complete data" (observed and estimated granule size by depth regression method) were used to develop two regression models for the granule size: an empirical model based on the process variables (inlet air temperature, inlet airflow rate, spray rate, and inlet air humidity) and a fundamental model based on the powder bed moisture content and the relative droplet size. The regression models based on the incomplete data from the previous study and those of the "complete data" were comparable in the sense that the contour plots and the predicted granule size were comparable.

   A.M. Van Nederkassel, A. Aerts, A. Dierick, D.L. Massart & Y. Vander Heyden.
"Fast separations on monolithic silica columns : method transfer, robustness and column ageing for some case studies". J. Pharm. Biomed. Anal., 32, 233-249 (2003).

Six separation methods, developed on conventional silica HPLC columns were transferred to monolithic silica columns. The transferred methods include the separation of an alkylbenzene mixture, of drugs from their impurities (nimesulide, tetracycline, phenoxymethylpenicillin and erythromycin) and of a green tea extract. The transfer of the first three methods was successful while for the latter three it was not. Increasine the flow rate up to 9 ml/min (where possible) decreased the analysis time of the transferred methods to 48 s (alkylbenzene mixture) 1.8 min (nimesulide mixture) and 3 min (tetracycline mixture) while still well separated peaks were obtained. The robustness and repeatability of these separations was found to be acceptable. Despite the use of flow rates up to 9 ml/min and frequent mobile phase changes the column performance was found rather constant and the column ageing minimal.

A. Nguyen Minh Nguyet, L. Tallieu, J. Plaizier-Vercammen, D.L. Massart & Y. Vander Heyden.

"Validation of an HPLC method on short columns to assay ketoconazole and formaldehyde in shampoo". J. Pharmac. Biomed. Anal., 32, 1-19 (2003).

An HPLC method to determine simultaneously ketoconazole and formaldehyde in an anti-dandruff shampoo, originally developed on a long column, was transferred to two short columns with similar stationary phase properties, but with a length of at the most 30% of the initial one. The fast HPLC methods on the short columns were validated. The validation characteristics consisted of selectivity, linearity range, precision (repeatability and time-different intermediate precision), bias and robustness. The assays for both ketoconazole and formaldehyde in the linear ranges showed no bias and an acceptable precision, although the latter was slightly worse than with the long one. The robustness tests were performed applying a Plackett- Burman design. For the ketoconazole assay, 6 factors were examined in a 12 experiments design and for formaldehyde, 11 factors in l6 experiments. The methods were found to be robust. The transfer seems to be successful and the obtained assays on the short columns are applicable for fast routine analysis.

   M. Daszykowski, B. Walczak & D.L. Massart.
"A journey into low-dimensional spaces with autoassociative neural networks". Talanta, 59, 1095-1105 (2003).

The compression and visualization of data has been always a subject of research. Since multidimensional data sets are difficult to interpret and visualize, much attention is paid to an efficient compression. Usually, the compression of dimensionality is the first step of exploratory data analysis. We focused our attention on autoassociative neural networks (ANNs), which provide data compression and visualization. ANNs can deal with linear and nonlinear correlation among variables, what makes them a very powerful tool in exploratory data analysis. ANNs are also often referred to as nonlinear principal component analysis (PCA), and due to their specific structure they are also known as bottleneck neural networks. In this paper, ANNs are discussed in details. Different training modes are described and illustrated on real examples. The usefulness of ANNs for nonlinear data compression and visualization purposes is proven with the aid of chemical data sets. The comparison of ANNs with PCA is also presented.

   Q.S. Xu, D.L. Massart, Y.-Z. Liang & K.-T. Fang.
"Two-step multivariate adaptive regression splines for modeling quantitative relationship between gas chromatography retention indices and molecular descriptors". J. Chromatography A, 998, 155-167 (2003).

The relationship between retention indices and molecular descriptors of alkanes is established by two-step multivariate adaptive regression splines (TMARS). TMARS combines linear regression with MARS. It is demonstrated that using linear regression or MARS modeling alone might cause lack of fit. TMARS avoids this and appreciably improves the prediction ability. The use of this combined approach permits the development of additional understanding of the adaptive nature in MARS modeling.

P. Dehouck, Y. Vander Heyden, J. Smeyers-Verbeke, D.L. Massart, J. Crommen, Ph. Hubert, R.D. Marini, O.S.N.M. Smeets, G. Decristoforo, W. Van de Wauw, J. De Beer, M.G. Quaglia, C. Stella, J.-L. Veuthey, O. Estevenon, A. Van Schepdael, E. Roets & J. Hoogmartens

"Determination of uncertainty in analytical measurements from collaborative study results on the analysis of a phenoxymethylpenicillin sample". Anal. Chim. Acta, 481, 261-272 (2003)

The correct interpretation of a measurement result requires knowledge about its uncertainty. Different operational definitions of uncertainty have been proposed, e.g. within-laboratory uncertainty, reproducibility uncertainty, bias-included uncertainty and absolute uncertainty. We considered the evaluation of reproducibility uncertainty derived from an interlaboratory experiment in which nine laboratories participated. The analyses consisted of a Karl-Fischer water determination, an acid-base titration to assay phenoxymethylpenicillin and a liquid chromatography (LC) method to determine 4-hydroxyphenoxymethylpenicillin and other impurities. The experimental set-up allowed to obtain for each determination estimates of the repeatability and the between-laboratory variance, respectively from which the reproducibility uncertainties were then derived.

B. Rambali, L. Baert & D.L. Massart

"Scaling up of the fluidized bed granulation process". Int. J. of Pharmaceutics, 252, 197-206 (2003)

The scaling up of a fluidized bed granulation process is often done empirically. Here, a systematic method was developed to achieve a similar granule size in the scaled up fluid bed. The scaling up is based on the relative droplet size, and the powder bed moisture content at the end of the spraying cycle. The present study describes the scaling up from small to medium and to production fluid bed scale. The granulation process is scaled up with as target a geometric mean granule size of 400 µm.

   J.A. Fernández Pierna. L. Jin, F. Wahl, N.M. Faber & D.L. Massart
"Estimation of partial least squares regression prediction uncertainty when the reference values carry a sizeable measurement error". Chem. Intel. Lab. Syst., 65, 281-291 (2003)

The prediction uncertainty is studied when using a multivariate PLSR model constructed with reference values containing a sizeable measurement error. Several expressions for calculating a sample-specific standard error of prediction have been proposed in the literature. In addition, Monte Carlo simulation methods, bootstrap and noise addition, can give an estimate of this uncertainty. In this paper, two approximate expressions are compared with the simulation methods for three near-infrared data sets.

   E. Hund, D.L. Massart & J. Smeyers-Verbeke
"Comparison of different approaches to estimate the uncertainty of a liquid chromatographic assay". Anal. Chim. Acta, 480, 39-52 (2003)

A measurement result cannot be properly interpreted without knowledge about its uncertainty. Several concepts to estimate this uncertainty have been developed. Four different approaches for uncertainty estimation are compared for an HPLC assay: the guide to the expression of uncertainty in measurement (GUM) approach; the top-down approach; an approach recently presented by Barwick and Ellison and finally an approach which estimates the measurement uncertainty from a robustness test. The comparison shows that the different approaches lead to comparable uncertainty estimates.

   R. Put, C. Perrin, F. Questier, D. Coomans, D.L. Massart & Y. Vander Heyden
"Classification and regression tree analysis for molecular descriptor selection and retention prediction in chromatographic quantitative structure - retention relationship studies". J. Chromatography A, 988, 261-276 (2003)

The use of the CART methodology was studied in a QSRR context on the retentions of 83 structurally diverse drugs. The response in the tree models consisted of the predicted retention factor (log k_w), while 266 molecular descriptors were used as explanatory variables. Molecular descriptors related to the hydrophohicity (log P and Hy) and the size (TPC) of the molecules were selected to describe and predict retention. CART was also able to select hydrogen-bonding and molecular complexity descriptors. Since these variables are expected from QSRR knowledge, it demonstrates the potential of CART as a methodology to understand retention in chromatographic systems. The potential of CART to predict retention for a given mixture was also evaluated. Reasonably good prediction, i.e. only 9% serious misclassifcation, was observed.

E. Van Gyseghem, S. Van Hemelryck, M. Daszykowski, F. Questier, D.L. Massart & Y. Vander Heyden

"Determining orthogonal chromatographic systems prior to the development of methods to characterise impurities in drug substances". J. Chromatography A, 988, 77-93 (2003)

As starting conditions for method development in LC the parallel application of generic orthogonal chromatographic systems could be useful. The possibilities to define orthogonal chromatographic systems were examined from correlation coefficients, by visual evaluation of selectivity differences, using principal component analysis, drawing color maps and evaluating dendrograms. Besides the stationary phase, the influence of pH and of organic modifier, on the selectivity of a system were evaluated. The dendrograms of hierarchical clusters were found good aids to assess orthogonality of chromatographic systems. The orthogonality of systems was validated using cross-validation and validation sets.

A. Detroyer, Y. Vander Heyden, I. Cambré & D.L. Massart.

"Chemometric comparison of recent chromatographic and electrophoretic methods in a quantitative structure-retention and retention-activity relationship context". J. Chromatography A, 986, 227-238 (2003).

The retention characteristics of basic pharmaceutical substances with a considerable difference in hydrophobicity were considered on an Immobilized Artificial Membrane column, with a Micellar Liquid Chromatography and a Micellar Electrokinetic Capillary Chromatography method. Applying Principle Component Analysis (PCA), it was seen the main retention principle was the same, however the above methods as well as more classical RP-HPLC methods varied in secondary retention mechanisms. When combining the results of the different methods a differentiation of the substances into their pharmacological families could be noticed with PCA. The high correlations of the retention characteristics with log P and a biological parameter seemed little affected by the method used.

Y. Vander Heyden, S.-T. Popovici, B.B.P. Staal & P.J. Schoenmakers

"Contribution of the polymer standards' polydispersity to the observed band broadening in size-exclusion chromatography". J. Chromatography A, 986, 1-15 (2003)

The contribution of the polydispersity of polymer standards to the observed band broadening in size-exclusion chromatography was evaluated. Theoretical predictions were found to overestimate this contribution, due to the fact that the polydispersity values specified by the manufacturers are too high to be applied in this context. An improved estimate of the polydispersity was obtained from the size-exclusion chromatography results and was used to reassess the polydispersity contribution to band broadening. When measering at optimal conditions, for low-molar-mass polymers, the estimated polydispersity index values approach the theoretical ones derived from a Poisson distribution.

J. Luypaert, M.H. Zhang & D.L. Massart

"Feasibility study for the use of near infrared spectroscopy in the qualitative and quantitative analysis of green tea, Camellia sinensis (L.)". Anal. Chim. Acta, 478, 303-312 (2003)

The possibility to use near infrared spectroscopy combined with PLS as a rapid method to estimate the quality of green tea is studied. NIR is used to build calibration models for caffeine, epigallocatechin gallate (EGCG) and epicatechin (EC) and for the total antioxidant capacity of green tea. To determine the total antioxidant capacity, the trolox equivalent antioxidant capacity (TEAC) method is used as reference method. For caffeine and TEAC, models are build for whole green tea leaves and for ground leaves; for the EGCG and EC, only for the whole leaves. A partial least squares algorithm is used for the calibration. To decide upon the number of PLS factors in the model, the lowest root mean square error of cross-validation for the training set is chosen. The correlation coefficient between predicted and reference results for the test set is used as an evaluation parameter for the models: for the TEAC results r = 0.90 for model with whole leaves, r = 0.86 for model with powdered leaves. The caffeine prediction model has an r = 0.96 for whole leaves and r = 0.93 for ground leaves. The correlation coefficient for the EGCG and the EC content models are 0.83 and 0.44 respectively.

S. Gourvénec, C. Lamotte, P. Pestiaux & D.L. Massart.

"Use of orthogonal projection approach (OPA) to monitor batch processes". Appl. Spectroscopy, 57 [1], 80-87 (2003)

The Orthogonal Projection Approach (OPA) and Multivariate Curve Resolution (MCR) are presented as a way to monitor batch processes using spectroscopic data. Curve resolution allows to look within a batch and to predict on-line real concentration profiles of the different species appearing during reactions. Taking into account the variations of the process by using an augmented matrix of complete batches, the procedure explained here calculates some prediction coefficients that can afterwards be applied for a new batch.

   W. Wu, Q. Guo, D.L. Massart, C. Boucon & S. de Jong.
"Structure preserving feature selection in PARAFAC using a genetic algorithm and Procrustes analysis". Chem. Intell. Lab. Syst., 65, 83-95 (2003)

A method is proposed to select subsets of variables in parallel factor analysis (PARAFAC), such that information in the complete multi-way data set is preserved. The information retained is measured by Procrustes analysis. The best N-way subset is obtained by applying a genetic algorithm (GA) to optimize the consensus between the subset and the complete N-way data set. The method was applied to two industrial data sets: a three-way sensory data set and a four-way gas chromatography (GC) data set. The results showed that the proposed method successfully identified structure-bearing variables in both data sets and that it led to better subsets of variables than feature selection based on loadings.

   L. Jin, J.A. Fernandez Pierna, F. Wahl, P. Dardenne & D.L. Massart.
"The law of mixtures method for multivariate calibration". Anal. Chim. Acta, 476, 73-84 (2003).

The Law of Mixtures (LM) method is a new topological method for multivariate calibration. It is shown to be a good method to predict for new objects that are inside the convex hull, determined by the calibration data. A method is also proposed for objects that are outside the convex hull.

   M. Daszykowski, B. Walckzak & D.L. Massart.
"Projection methods in chemistry". Chem. Intel. Lab. Syst., 65, 97-112 (2003).

Visualization of a data set structure is a challenging goal in data mining. Chemical data sets are often multidimensional, and visualization of their structure is not directly possible. To overcome this problem, the original data is compressed by using projection techniques, preserving the original data structure, and allowing its visualization. In this paper, a survey of different projection techniques, linear and nonlinear, is given. Their performance is illustrated on chemical data sets.

   M. Daszykowski, B. Walckzak & D.L. Massart.
"On the optimal partitioning of data with K-means, neural gas and growing neural gas". J. Chem. Inf. Sci., 42, 1378-1389 (2003)

The performance of new clustering methods such as Neural Gas (NG) and Growing Neural Gas (GNG) is compared with the K-means method for real and simulated data sets. A new algorithm, growing K-means, GK, is introduced as the alternative to Neural Gas and Growing Neural Gas. It has small input requirements and is conceptually very simple. The GK leads to nearly optimal values of the cost function, and, contrary to K-means, it is independent of the initial data set partition. The incremental property of GK additionally helps to estimate the number of "natural" clusters in data.

D. Visky, Y. Vander Heyden, T. Iványi, P. Baten, J. De Beer, Z. Kovács, B. Noszal, E. Roets, D.L. Massart & J. Hoogmartens.

"Characterisation of reversed phase liquid chromatographic columns by chromatographic tests. Evaluation of 36 test parameters : repeatability, reproducibility and correlation". J. Chromatography A, 977, 39-58 (2002).

Official compendia only give a general description of the stationary phase in the description of a liquid chromatographic method. Therefore the selection of a column with suitable selectivity might present difficulties. Earlier, a test procedure was proposed for 36 parameters reported to be representative for stationary phase characteristics. This paper describes the application of the test procedure on 69 RP-LC C₁₈ columns. The chromatographic parameters obtained were evaluated, and their repeatability, reproducibility and correlation examined.

D. Ameye, E. Keleb, C. Vervaet, J.P. Remon, E. Adams and D.L. Massart.

"Scaling-up of a lactose wet granulation process in Mi-Pro high shear mixers". European Journal of Pharmaceutical Sciences, 17, 247-251 (2002).

The high shear wet granulation upscaling possibilities of a Pro-C-epT Mi-Pro high shear mixer were investigated. The wet granulation process was first optimised in the Gral 10 (8 l batch size), then an octagonal design was build around the central point. Two process parameters, the impeller tip speed and the water content, were varied at three levels, in an experimental design. α-Lactose monohydrate was granulated using a polyvinylpyrollidone binder solution. The granules were dried, sieved, characterised and compressed to tablets. In all mixer volumes examined the used impeller tip speed range did not influence the granule or tablet properties. In all bowl volumes the influence of water concentration on actual yield, particle size distribution and granule friability was similar. All batch sizes resulted in tablets of similar quality. For the selected formulation the lab scale Mi-Pro high shear mixers with different bowl volumes could be used to determine the optimal process parameters and to scale up to pilot scale.

   E. Hund, Y. Vander Heyden, D.L. Massart & J. Smeyers-Verbeke.
"Derivation of system suitability test limits from a robustness test on an PLC assay with complex antibiotic samples". J. Pharm. Biomed. Anal., 30, 1197-1206 (2002).

A System Suitability Test (SST) verifies the adequate working of the equipment for analytical measurements. In pharmaceutical analysis, SSTs are performed at least at the beginning of a series of routine analyses. The most generally applied SST considers the precision of the analysis, i.e. the repeatability. Additionally, a SST can also consider responses indicative for the quality of the technique used, e.g. resolutions or peak asymmetry in LC. The system is declared suitable if the response is within given limits. However, it is not always evident how to define the SST limits for a newly developed method. Robustness tests have been proposed as a starting point to deduce limits. Here, it is examined how such a strategy can be applied for complex samples of microbial origin.

   M. Daszykowski, B. Walczak & D.L. Massart.
"Representative subset selection". Anal. Chim. Acta, 468, 91-103 (2002).

Nowadays analytical techniques enable to acquire huge amount of data. Large data sets are difficult to handle and a big interest exists in designing a subset of the original data set, which preserves the information and facilitates the computations. Many subset selection methods are described and their choice depends on the problem. The two most popular subset selection methods are uniform and cluster-based designs. In this paper uniform designs, such as those proposed by Kennard and Stone, OptiSim, and cluster-based designs applying K-means technique and density based spatial clustering of applications with noise (DBSCAN) are considered. Additionally, a new concept of the subset selection with K-means is introduced.

   J. Luypaert, S. Heuerding, S. de Jong & D.L. Massart.
"An evaluation of direct orthogonal signal correction and other preprocessing methods for the classification of clinical study lots of a dermatological cream". J. Pharmac. Biomed. Anal., 30, 453-466 (2002).

Different preprocessing methods (direct orthogonal signal correction, standard normal variate, multiple scatter correction, first and second derivation, offset correction and detrend correction) are applied to NIR spectra of dermatological creams with different concentrations active compound. The influence of the preprocessing methods on the classification of the samples in its concentration class is evaluated using nearest neighbour methods (with Euclidean distance and correlation coefficient as distance parameters) as classification method. PLS and PCR modelling are also used to predict the concentration active compound. The direct orthogonal signal correction gave best results.

M. Praisler, J. Van Bocxlaer, A. De Leenheer & D.L. Massart.

"Chemometric detection of thermally degraded samples in the analysis of drugs of abuse with gas chromatography-Fouriertransform infrared spectroscopy". J. Chromatography A, 962, 161-173 (2002).

A chemometric procedure to identify the reference standard chromatographic peak in GC - FTIR analysis of commercial standards with more than one peak in the GC chromatogram is presented. The procedure has been designed for phenethylamines, the largest class of illicit drugs, which are abused for their stimulant and/or hallucinogenic effects. The similarity between their vapor-phase FTIR spectra was modeled using principal component analysis, and class identity was assigned based on soft independent modeling of class analogy.

F. Questier, B. Walczak & D.L. Massart.

"Feature selection for hierarchical clustering". Anal. Chim. Acta, 466, 311-324 (2002).

Feature selection is a technique in data analysis for information-preserving data reduction. This paper describes a feature selection approach for hierarchical clustering based on genetic algorithms. Clustering trees based on reduced feature sets are comparable with those based on the complete set. Special measures to favor small reduced feature sets are discussed.

S. Hillaert, Y. Vander Heyden and W. Van den Bossche.

"Optimisation by experimental design of a capillary electrophoretic method for the separation of several inhibitors of the angiotensin-converting enzyme using alkylsulphonates". Journal of Chromatography A, 978, 231-242 (2002)

An experimental design was used to optimise a capillary electrophoretic method for eight inhibitors of the angiotensin-converting enzyme. The usefulness of alkylsulphonates as ion-pairing agents was investigated. After preliminary investigations a three-level full-factorial design was applied to study the impact of the pH and the molarity of the ion-pairing agent on the separation. Improved separations were obtained suggesting a favourable effect of ion-pairing interactions between analytes and additive. A combination of two systems was still necessary for the selective identification of all substances.

H.D. Spapen, D.N. Nguyen, K. ver Elst, Y. Vander Heyden and L.P. Huyghens.

"N-acetylcysteine and cardiac disfunction in human septic shock", Clinical Intensive Care Journal, 13, 1, 27-32 (2002)

The objective was to determine which patients (N:12) in septic shock may develop myocardial depression during adjuvant therapy with N-acetylcysteine (NAC). It was found that NAC metabolism during septic shock is disturbed in patients with underlying cardiac disease. The resulting high plasma NAC concentrations may expose these patients to adverse cardiovascular events. Increased plasma cardiac troponin I levels in association with left ventricular dysfunction as assessed by transoesophageal echocardiography may identify patients "at risk".

I. Ruisánchez, F.X. Rius, S. Maspoch, J. Coello, T. Azzouz, R. Tauler, L. Sarabia, M.C. Ortiz, J.A. Fernández, D. Massart, A. Puigdomčnech & C. García.

"Preliminary results of an interlaboratory study of chemometric software and methods on NIR data. Predicting the content of crude protein and water in forages." Chemometrics and Intelligent Laboratory Systems, 63, 93-105 (2002).

Within the European Commission project 'European Network for the Intercomparison of Chemometric Software and Methods', a proficiency study on a near infrared spectroscopy (NIR) data set was carried out to determine how different the results from six laboratories were when different multivariate calibration methods and software were used. All participants had previous experience on mulvariate calibration. The data consisted of NIR spectra of 305 forage samples recorded under the same conditions. Two parameters were predicted, moisture at 103-105 °C and crude protein content. Results showed that the root mean square error of prediction (RMSEP) values obtained by all laboratories were acceptable, alhough they varied considerably.

   V. Pravdova, C. Boucon, S. de Jong, B. Walczak & D.L. Massart.
"Three-way principal component analysis applied to food analysis : an example." Anal. Chem. Acta, 462, 133-148 (2002)

The purpose of the study is to show how the interpretation of a complex multivariate data array can be improved using N-way principal component analysis. Two food related three-way data sets were studied; a sensory and a chromatographic one. The Parafac and the Tucker3 models were applied and results compared. Both N-way models presented allow visualization of the data structure and give detailed information about the data set, allowing to understand relationships between objects and variables.

   F. Questier, I. Arnaut-Rollier, B. Walczak & D.L. Massart.
"Application of rough set theory to feature selection for unsupervised clustering". Chem. Int. Lab. Systems, 63, 155-167 (2002)

Feature selection is a valuable technique for information-preserving data reduction. This paper describes the use of rough set theory (RST) for reducing in a supervised way the number of features in unsupervised clustering. The application to a hierarchical clustering of Pseudomonas species is presented. The Wallace measure is used for the comparison of the clustering results based on the original and on the reduced data set.

   J.A. Fernández Pierna, F. Wahl, O.E. de Noord and D.L. Massart.
"Methods for outlier detection in prediction". Chemometrics and Intelligent Laboratory Systems, 63, 27-39 (2002).

If a prediction sample is different from the calibration samples, it can be considered as an outlier in prediction. Two techniques, uncertainty estimation and convex hull method are studied to detect such outliers. Classical techniques (Mahalanobis distance and X-residuals), potential functions and robust techniques are used for comparison. The combination of the convex hull and the uncertainty estimation offers a practical way for detecting outliers in prediction. Inliers can also be detected by adding the potential function method.

   E. Hund, D.L. Massart and J. Smeyers-Verbeke.
"Robust regression and outlier detection in the evaluation of robustness tests with different experimental designs". Anal. Chim. Acta, 463, 53-73 (2002).

Robustness tests are usually based on experimental designs. As such experiments generally lead to a large variability in the results, erroneous results are often not detected. As a consequence, the ordinary least squares (OLS) estimates of the effects can be biased. Two robustness tests which contain a suspicious result, are studied. Moreover, simulated datasets examine the influence of the extent of the outlier and of multiple outliers. First, different methods are applied to inspect the results for outliers: the half-normal plot of the OLS residuals, the normal probability plot of the effects and a method based on experimental design reconstruction. Secondly, two robust regression methods are applied to calculate the effects with a minimum influence of possible outliers. The methods are compared and it is evaluated when they can be applied.

S. Lauwers, Y. Vander Heyden and B. Rombaut.

"Screening and optimisation of an ELISA method for the quantitative detection of enterovirus specific RT-PCR products by means of a two-level experimental design". Journal of Pharmaceutical and Biomedical Analysis, 29, 659-668 (2002).

Earlier, optimal reaction conditions were determined for the RT-PCR part of a quantitative enterovirus specific RT-PCR ELISA method. To obtain a detection limit as low as possible, the ELISA part of the procedure was optimised . This was done by investigating seven factors at three levels in a multivariate approach. A reflected two-level screening design was used. Optimal reaction conditions were established from the effects of the factors on the measured absorbance of the ELISA detection. The linear range and detection limit of the test were determined and compared with those before optimisation.

   C. Perrin, N. Matthijs, D. Mangelings, C. Granier-Loyaux, M. Maftouh, D.L. Massart and Y. Vander Heyden.
"Screening approach for chiral separation of pharmaceuticals. Part II. Reversed phase liquid chromatography"; Journal of Chromatography A, 966, 119-134 (2002).

A screening strategy for the rapid separation of drug enantiomers was developed using three cellulose/amylose stationary phases. The key point for enantioselectivity is to control compound ionisation. Only two mobile phases, i.e. an acidic phosphate buffer (pH 2.0) containing a chaotropic salt (KPF₆) and a borate buffer (pH 9.0) mixed with acetonitrile, are used in the strategy. This strategy was successfully applied to a set of 37 diverse pharmaceuticals. Enantioselectivity was achieved for 89%.

Y. Vander Heyden, A. Nguyen Minh Nguyet, M.R. Detaevernier, D.L. Massart and J. Plaizier-Vercammen.

"Simultaneous determination of ketoconazole and formaldehyde in a shampoo : HPLC method development and validation". Journal of Chromatography A, 958, 191-201 (2002).

Ketoconazole is the active ingredient in a shampoo used to treat seborrhatic dermatitis. The shampoo also contains imidazolidinylurea as a formaldehyde releasing preservative. The aim was to develop an HPLC system to the determine both ketoconazole and formaldehyde. The isocratic system consisted of an Interchrom Nucleosil (250x4.6mm, 5µm) C8 column and a mobile phase containing acetonitrile - phosphate buffer 0.025M, pH 4.0, 45/55 (V/V). Ketoconazole could be determined at 250nm after injection of diluted shampoo. Formaldehyde was measured at 345nm after derivatisation with 2,4- dinitrophenylhydrazine. At the selected conditions, the other excipients did not interfere in the assays. Method validation was performed . Different selectivity towards ketoconazole and formaldehyde was observed when applying other C8 columns. This fact, however, did not affect the assays of both substances.

D. Visky, Y. Vander Heyden, T. Iványi, P. Baeten, J. De Beer, B. Noszal, E. Roets, D.L. Massart and J. Hoogmartens.

"Characterisation of reversed phase liquid chromatographic columns by chromatographic tests. Preliminary experiments and development of the protocol"; Pharmeuropa, 14.2, 288-297 (2002).

In the description of a liquid chromatographic method official compendia do not mention the brand of stationary phase to be used to obtain sufficient selectivity. Column selection presents difficulties since more than 600 RP-LC columns are available on the market. This project was started to determine chromatographic parameters that would enable the selection of an appropriate column. First a suitable test procedure is needed to measure a number of parameters representative for different groups of stationary phases. Test methods described in the literature for characterisation of RP-LC columns are reviewed. None of the described methods is sufficient to characterise all stationary phase properties. Therefore a combination of tests is proposed. The final test procedure was tried out on 4 columns.

Y. Vander Heyden, S.T. Popovici and P.J. Schoenmakers.

"Evaluation of size exclusion chromatography and size exclusion electrochromatography calibration curves". Journal of Chromatography A, 957, 127-137, (2002).

Size-exclusion chromatography (SEC) and size-exclusion electrochromatography (SEEC) are chromatographic techniques to determine molecular weight distributions (MWD) of polymers. One important step to derive MWD parameters is the modelling of the calibration curves, which was examined here. The calibration curves in SEC and SEEC are generally not linear. Different polynomial models were evaluated and compared, both for model fit and predictive properties. Sometimes a straight line and sometimes a third-order polynomial model were best. The best model across the effective or linear range is not always a straight line. The SEEC curves had considerably higher prediction errors than the SEC ones. Reduction of the number of calibration standards to five or six did not greatly affect the predictive properties of the calibration curves, neither in SEC nor in SEEC.

   E. Adams, D. Coomans, J. Smeyers-Verbeke & D.L. Massart.
"Non-linear mixed effects models for the evaluation of dissolution profiles". Int. J. Pharm., 240, 37-53 (2002)

The use of non-linear mixed effects models to describe dissolution data has been evaluated. The standard settings of the statistical software package (S-plus) are used . Several mathematical functions,e.g. Weibull, logistic, first-order and Gompertz, are employed as a basis for the non-linear mixed effects models. Examples using dissolution data of different types of tablets are given.The results are compared with those obtained using linear mixed effects models.

   W. Wu, Q. Guo, S. de Jong & D.L. Massart.
"Randomisation test for the number of dimensions of the group average space in generalised Procrustes analysis ". Food Quality & Pref., 13, 191-200 (2002)

When generalised Procrustes analysis (GPA) is applied to analyse sensory data, it is important to decide on the dimensionality in the final group average space. Usually, it is empirically estimated based on the scree plot of the total explained variance and on interpretability of dimensions. No statistical method is available so far. In this paper, a new randomisation test was developed to detect significant dimensions of group average space obtained by GPA. The method was applied to sensory and simulated data sets. The proposed randomisation F-test is found more effective than the scree plot method. It succesfully interpreted the different data sets.

   V. Schoonjans, N.Taylor, B.D. Hudson and D.L. Massart.
"Characterization of the similarity of chemical compounds using electrospray ionisation mass spectrometry and multivariate exploratory techniques". J. Pharmac. Biomed. Anal., 28, 537-548 (2002)

The applicability and usefulness of electrospray mass spectrometry to the analysis of a data set of 52 substances to probe their molecular similarity/diversity has been examined. The first stage was the reduction of the data by PCA to visualize the data structure. Sequential Projection Pursuit was applied to detect outlying objects and Hierarchical cluster analysis using group-average linkage clustering to produce a dendrogram. Finally, the cluster results of spectral data were compared with structural fingerprints and an expert's classification.

   Y. Vander Heyden, V. Pravdova, F. Questier, L. Tallieu, A. Scott and D.L. Massart.
"Parallel co-ordinate geometry and principal component analysis for the interpretation of large multi-response experimental designs ". Anal. Chim. Acta, 458, 397-415 (2002)

The use of parallel co-ordinate geometry (PCG) plots, PCA and N-way PCA as visualization procedures for large multi-response designs was compared with the more traditional approach of calculating factor effects by linear regression. PCG plots are a recent addition to the visualization tools and offer a possibility to visualize multi-dimentional data in two dimensions. It was found that PCA and PCG each have their own benefits and disadvantages. Both methods can be used to some extent to select optimal conditions. The PCA score plots could be used as a Pareto optimality plot. The examined visualizition methods can be useful and complementary aids in the interpretation of large multi-response experimental design data and they add a multivariate dimension to the more classical univariate analysis of such data.

   M. Daszykowski, B. Walczak and D.L. Massart.
"Looking for natural patterns in analytical data. Part 2. Tracing local density with OPTICS". J. Chem. Inf. Comput. Sci., 42, 500-507 (2002).

The principles and the algorithm of a density-based clustering approach, OPTICS, are described, and its properties, i.e. ability to reveal clusters and different densities, are illustrated on spectral and chromatographic data sets. A "reachability plot" visualizing density fluctuations of data in multivariate space and a "color map" relating the original and/or descriptive features with data clustering allow a deeper insight into the data structure and better chemical interpretation.

T. Ivanyi, Y. Vander Heyden, D. Visky, P. Baten, J. De Beer, I. Lázár, D.L. Massart, E. Roets and J. Hoogmartens.

"Minimal number of chromatographic test parameters for the characterisation of reversed-phase liquid chromatographic stationary phases". J. Chromatography A, 954, 99-114 (2002).

This paper focuses on the classification of RP-HPLC columns based on chromatographic properties. A chemometric study has been conducted on data sets consisting of 85 columns and of 47 self-tested columns. Principal component analysis enables determination of the number of parameters necessary for a differentiation. The results show that reducing the number of parameters still allows classification of the columns . It is shown that three test parameters produce a classification similar to that obtained with five.

   V. Pravdova, B. Walczak and D.L. Massart.
"A comparison of two algorithms for warping of analytical signals". Anal. Chim. Acta, 456, 77-92 (2002)

The alignment of analytical signals is an important preprocessing step when further analysis (e.g. PCA) requires the same lengths of all of them. Two algorithms for alignment of profiles, namely dynamic time warping (DTW) and correlation optimized warping (COW) were tested and compared. The special attention was focused on chromatographic and spectroscopic profiles. Simulated and two sets of real data were studied in this study.

   C. Vannecke, M.S. Bloomfield, Y. Vander Heyden and D.L. Massart.
"Development of a generic tlow-injection analysis method for compounds with a secondary amine or amide function using an experimental design approach. Part II. Selection and evaluation of the chemical reaction parameters". Anal. Chim. Acta, 455, 117-130 (2002)

This part consists in the selection and evaluation of the chemical reaction conditions. Sodium hypochlorithe first converts the secondary amine or the amide to a primary amine. The latter reacts with o-phthalaldehyde (OPA) and a thiol (N-acetylcysteine (NAC)) to form a derivative which can be measured fluorimetrically. To investigate the influence of the different chemical reaction parameters on the peak height for a set of 31 pharmaceutical compounds, a quarter-fraction factorial design for six factors at two levels (2^6-2-resolution IV, 16 experiments) was executed. Effects on the responses were calculated for each compound.

   Q. Guo, W. Wu, D.L. Massart, C. Boucon and S. de Jong.
"Feature selection in principal component analysis of data". Chemometrics and Intelligent Laboratory Systems, 61, 123-132 (2002)

A feature selection method is proposed to select a subset of variables in principal component analysis (PCA) that preserves as much information present in the complete data as possible. The information is measured by means of the percentage of consensus in generalised Procrustes analysis. The best subset of variables is obtained by applying a genetic algorithm (GA) to optimise the consensus between the subset and the complete data set. The method was evaluated on a standard data set known and on a high-dimensional industrial gas chromatography (GC) data set. The results showed that the proposed method leads to a better subset of variables than other studied feature selection methods.

   S. Gourvénec, D.L. Massart and D.N. Rutledge.
"Determination of the number of components during mixture analysis using the Durbin-Watson criterion in the Orthogonal Projection Approach and in the SIMPLe-to-use Interactive Self-modelling Mixture Analysis approach". Chem. Int. Lab. Syst., 61, 51-61 (2002)

The Orthogonal Projection Approach (OPA) and the SIMPLe-to-use Interactive Self-modelling Mixture Analysis approach (SIMPLISMA) are employed during process monitoring to obtain concentration profiles and/or pure spectra of a mixture. In the first step of these methods, it is important to select the right number of components. The Durbin - Watson criterion was applied to dissimilarity values in OPA and to purity vaiues in SIMPLISMA as a tool for the decision of the number of components. This yields more objective results than visual interpretation.

   F. Questier, Q. Guo, B. Walczak, D.L. Massart, C. Boucon and S. de Jong.
"The neural-gas network for classifying analytical data". Chemometrics and Intelligent Laboratory Systems, 61, 105-121 (2002)

This article introduces the Neural-Gas network, a fast neural net-based method for clustering, and shows how it is applied to gas chromatographic patterns of Maillard reaction products. The advantages of Neural-Gas are compared to the K-means clustering method and to the Kohonen self-organising maps. Novel combinations with visualization techniques are also presented.

   E. Adams, B. Walczak, C. Vervaet, P.G. Risha and D.L. Massart.
"Principal component analysis of dissolution data with missing elements ". Int. J. Pharmaceut., 234, 169-178 (2002)

The use of principal component analysis (PCA) for incomplete dissolution data sets is examined. The PC space is constructed using a reference set and the test set is projected in that space. Several cases such as a reference set with missing data, an incomplete test set and both sets measured at different time points, are discussed using two examples: one simulation and one obtained from the pharmaceutical practice. From the many possibilities to deal with missing data, the expectation - maximization algorithm in combination with PCA was chosen.

   C. Perrin, V.A. Vu, N. Matthijs, M. Maftouh, D.L. Massart and Y. Vander Heyden.
"Screening approach for chiral separation of pharmaceuticals. Part I: normal phase liquid chromatography". J. Chromatogr. A, 947, 69-83 (2002)

A strategy for rapid screening for the separation of chiral molecules of pharmaceutical interest by normal-phase liquid chromatngraphy using three cellulose/amylose stationary phases is proposed. In a first step, the most important parameters for the separations were determined and studied for their effects by means of experimental designs. Results showed that the cellulose tris-(3,5-dimethylphenylcarbamate), the amylose tris-(3,5-dimethylphenylcarbamate) and the cellulose tris-(4-methylbenzoate) stationary phases have very broad and complementary enantiorecognition properties. The type of organic modifier used in the mobile phase appeared to have a dramatic influence on the quality of the separation.

M. Praisler, J. Van Bocxlaer, A. De Leenheer and D.L. Massart.

"Automated recognition of ergogenic aids using Soft Independent Modeling of Class Analogy (SIMCA)". Turkish J. Chem., 26, 45-58 (2002)

The introduction of effective drug testing procedures in doping control reduced, but did not eliminate, their abuse by athletes. The most analytical challenge is the recognition of the new analog compounds. In attempts to circumvent existing controlled substance laws, slightly modified chemical structures (by adding or changing substituents at various positions on the banned molecules) are used. As a result, no substance belonging to a prohibited class may be used nowadays, even if it has not been specifically listed. We present a chemometric procedure acting as an automated GC-FTIR screening test detecting the molecular structural similarity of unknown compounds with the main classes of ergogenic aids. The class identity of a compound is diagnosed, using Soft Independent Modeling of Class Analogy (SIMCA). The specificity and selectivity of the screening test, evaluated by testing 160 toxicological relevant compunds, are discussed.

N. Matthijs, C. Perrin, M. Maftouh, D.L. Massart and Y. Vander Heyden.

"Knowledge-based system for method development of chiral separations with capillary electrophoresis using highly-sulfated cyclodextrins". J. Pharmac. Biomed. Anal., 27, 515-529 (2002)

Method development for chiral separations requires experience. In order to help the analyst, a knowledge-based system (KBS) for the rapid determination of experimental parameters, which allow a baseline separation of enantiomers, has heen developed. Rules were defined and a KBS was built. This paper describes use of capillary electrophoresis (CE), in which a strategy has heen defined based on the use of highly-sulfated cyclodextrins as chiral selectors. A structured representation of the knowledge and its implementation in Toolbook software is presented.

P.J. Rousseeuw, S. Van Aelst, B. Rambali and J. Smeyers-Verbeke.

"Deepest regression in analytical chemistry". Anal. Chim. Acta, 446, 245-256 (2001)

V. Simeonov, G. Andreev, D.L. Massart and S. Tsakovski.

"Environmental interpretation of analytical data of marine organisms from the Black Sea". Toxicological and Environmental Chemistry, 80, 53-65 (2001)

The paper deals with the application of cluster analysis and principal components analysis to estimate the reaction specificity (bioindication ability) of some benthic organisms from the Black Sea. The bioindication of marine organisms and algae with respect to heavy metal pollution of the marine environment proves to be a very important step in risk assessment and quality control procedures. In the study it is indicated that the multivariate statistical treatment of monitoring data sets for heavy metals concentrations in various benthic organisms and algae could help in establishing the bioindicative ability of the species.

R. Dams, T. Benijts, W.E. Lambert, D.L. Massart and A.P. De Leenheer.

"Heroin impurity profiling : trends throughout a decade of experimenting". Forensic Science International, 123, 81-88 (2001)

Heroin is still one of the most frequently abused drugs of today. All over the world, law enforcement agencies try to eradicate the illicit production and trafficking of this potent and highly addictive narcotic. To this aim, important information is provided by physical and chemical toxicological analysis of confiscated samples, with special attention for the identification and the quantification of minor components, such as the impurities related to the origin and manufacturing. By combining these data complex characterisations, i.e. impurity profiles, chemical signatures or fingerprints, can be obtained and used for comparative analysis

   Q. Guo, W. Wu, D.L. Massart, C. Boucon and S. de Jong.
"Feature selection in sequential projection pursuit". Anal. Chim. Acta, 446, 85-96 (2001)

A feature selection method is proposed to select a subset of variables in sequential projection pursuit (SPP) analysis in order to preserve as much sample clustering information as possible. The inhomogeneity of the complete data is explored by SPP, and the retained inhomogeneity information of a candidate subset is measured by means of the percentage of consensus in generalised procrustes analysis. The best subset is obtained by applying a genetic algorithm (GA) which optimises the consensus between the subset and the complete data set.

B. Chimanuka, G. François, G. Timperman, Y. Vander Heyden, J. Holenz, J. Plaizier-Vercammen, G. Bringmann.

"A comparison of the stage-specific efficacy of chloroquine, artemether and dioncophylline B, against the rodent malaria parasite Plasmodium chaboudi chaboudi in vivo". Parasitology Research, 87, 795-803 (2001).

Chloroquine, artemether and dioncophylline B efficacy against Plasmodium chabaudi was compared. One intraperitoneal injection was given daily over 3 consecutive days to mice when they were bearing ring, trophozoite and schizont forms. The parasitaemia was monitored until parasites were cleared. Chloroquine was more efficient at the trophozoite stage, while artemether was effective against all erythrocytic stages. Chloroquine-treated and artemether-treated parasites lowered the parasitaemia faster than dioncophylline B. Dioncophylline B was effective at trophozoite and schizont stages, but ineffective at ring stage. A better timing of drug administration increases the efficacy of antimalarial drugs.

   C. Vannecke, E. Van Gyseghem, M.S. Bloomfield, T. Coomber, Y. Vander Heyden and D.L. Massart.
"Development of a generic FIA-method for compounds with a secondary amine or amide function, using an experimental design approach - Part I : Selection and evaluation of the FIA-system parameters" Analytica Chimica Acta, 446, 413-428 (2001)

A generic flow injection analysis (FIA) method to determine compounds with a secondary amine or an amide in their structure was developed. A first part of the development, namely the selection and evaluation of the FIA-system, is described. To investigate the influence of the different FIA-system parameters on both the peak height and the residence time for a set of twenty-eight pharmaceutical compounds, a half-fraction factorial design for five factors at two levels was executed. Effects on both responses were calculated. Since the peak height should be maximized and the residence time minimized, the Pareto optimality concept was applied to make an optimal compromise between both responses and to eventually select a generic FIA-system.

   F. Estienne, L. Pasti, V. Centner, B. Walczak, F. Despagne, D. Jouan Rimbaud, O.E. de Noord and D.L. Massart.
"A Comparison of Multivariate Calibration Techniques Applied to Experimental NIR Data Sets. Part II : Predictive Ability under Extrapolation Conditions" Chemometrics and Intelligent Laboratory Systems, 58, 195-211 (2001)

In this study, performances of different multivariate calibration techniques are compared in extrapolation conditions (predicted sample fall outside the calibration domain). Different types and levels of extrapolation are considered. The study is performed on five near-IR data sets representative of difficulties met in practical cases (non-linearity, non-homogeneity and presence of irrelevant variables in the set of predictors). General recommendations are made about what method to use when samples requiring extrapolation can be expected.

   F. Estienne and D.L. Massart.
"Multivariate calibration with Raman data using fast principal component regression and partial least squares regression methods". Anal. Chim. Acta, 450, 123-129 (2001)

Linear and non-linear calibration methods (principal component regression (PCR), partial least squares regression (PLS), and neural networks (NN)) were applied to a slightly non-linear Raman data set. Because of the large size of this data set, recently introduced linear calibration methods, specifically optimised for speed, were also used. These fast methods achieve speed improvement by using the Lanczos decomposition for the singular value decomposition steps of the calibration procedures, and for some of their variants, by optimising the models without cross-validation (CV). Linear methods could deal with the slight non-linearity present in the data by including extra components, therefore, performing comparably to NNs.

V. Simeonov, C. Sarbu, D.L. Massart and S. Tsakovski.

"Danube River Water Data Modelling by Multivariate Data Analysis". Mikochimica Acta, 137, 243-248 (2001)

A data set (48 x 19) consisting of river water analytical data during a four-year period has been treated by principal components analysis (PCA). The PCA indicated that seven latent factors ("hardness", "biochemica", "waste inlets", "turbidity", "acidity", "soil extracts" and "organic wastes") are responsible for the data structure. Its complexity is further proved by the application of multiple linear regression analysis on the absolute principal components scores (APCS) where the contribution of each natural or anthropogenic sources in the factor formation is shown.

V. Pravdova, B. Walczak, D.L. Massart, S. Kawano, K. Toyoda and R. Tsenkova.

"Calibration of somatic cell count in milk based on near-infrared spectroscopy". Anal. Chim. Acta, 450, 131-141 (2001)

Acceptable PLS model for prediction of somatic cell count (SCC) based on NIR spectra of unhomogenized milk is presented in the study. Samples of raw milk were collected from cows in the early lactation period (from 7th to 29th day after parturition). The NIR spectra were measured in the region 400-1100 nm. As reference method a fluoro-opto-electronic method was applied. Different preprocessing methods were investigated. The robust version of PLS regression was applied to handle outliers present in the data set and the UVE-PLS method was used to eliminate uninformative variables.

   V. Pravdova, F. Estienne, B. Walczak and D.L. Massart.
"A robust version of the Tucker3 model". Chemom. Intell. Lab. Systems, 59, 75-88 (2001)

A new procedure for identification of outliers in Tucker3 model based on robust initialization of the Tucker3 algorithm using Multivariate trimming or Minimum covariance determinant is proposed.The performance of the algorithm is tested by a Monte Carlo study on simulated data sets and also on a real data set known to contain outliers.

   C. Perrin, B. Walczak and D.L. Massart.
"The use of wavelets for signal denoising in capillary electrophoresis". Anal. Chem., 73, 4903-4917 (2001)

The discrete wavelet transform was applied to denoise electropherograms in capillary electrophoresis (CE). The use of the Haar wavelet and translation invariant denoising were found to be very efficient for this purpose. An important improvement was obtained, as compared with Savitzky - Golay and Fourier, which are the most commonly used techniques for denoising in the instrumentation software packages. A better removal of the noise and, especially, a better preservation of the shapes of very sharp peaks was achieved. Removal of the baseline variations was also investigated.

   B. Walczak, et al
"Dealing with missing data Part II". Chemom. Intell. Lab. Systems, 58, 29-42 (2001)

The main concepts of the maximum likelihood (ML) approach in dealing with missing data are introduced and simple numerical examples of the application of ML are presented. Differences between ML and other techniques of treating missing data are illustrated. The idea of multiple imputation (MI) approach is presented and illustrated as well.

   B. Walczak and D.L. Massart
"Dealing with missing data Part I". Chemom. Intell. Lab. Systems, 58, 15-27 (2001)

The principles of the iterative approach to deal with missing data are presented and the performance of this approach in multivariate methods such as PCA, PCR, PLS and N-way is studied. Matlab codes for these analyses are appended.

   F. Estienne, N. Matthijs, D.L. Massart, P. Ricoux and D. Leibovici.
"Multi-way modelling of high-dimensionality electroencephalographic data". Chem. Intel. Lab. Syst., 58, 59-72 (2001)

In this study there is investigated whether useful information can be extracted from an electroencephalographic (EEG) data set with a very high number of modes, and there is also determined which model is the most appropriate for this purpose. The data is acquired of the testing phase of a new drug expected to have an effect on the brain activity. The implemented test program led to a six-way data set. It was verified that a multidimensional structure was present and therefore multi-way methods were used to model the data. It appeared that Tucker 3 was the most suited model. It was possible to extract useful information from this high-dimensionality data.

   E. Hund, D.L. Massart and J. Smeyers-Verbeke.
"Operational definitions of uncertainty". TrAC, vol.20, no.8, 394-406 (2001)

Very different approaches for the estimation of the uncertainty related to measurement results are found in the literature and in published guidelines. This article analyses and compares them. 'Uncertainty' is not, and should not be, the same in all situations. As a consequence, operational definitions of uncertainty are proposed that take into account the differences in the ways in which truth, uncertainty and error are conceived.

B. Rambali, J.A. Fernández, L. Van Nuffel, F. Woestenborghs, L. Baert, D.L. Massart and F.C. Odds

"Susceptibility testing of pathogenic fungi with itraconazole: a process analysis of test variables". J. Antimicrobial Chem. 48, 163-177 (2001)

   S. Kuttatharmmakul, D.L. Massart, D. Coomans and J. Smeyers-Verbeke.
"Comparison of methods for the estimation of statistical parameters of censored data". Anal. Chim. Acta, 441, 215-229 (2001)

Approaches based on the maximum likelihood (ML) method and on the order statistics are described and evaluated for the estimation of the mean and standard deviation of a normal population from a left-singly censored sample, i.e. a sample for which some measurement results fall below the reporting limit of the analytical method. The performance of the methods is evaluated by means of data simulatians. The sample size considered is small to moderate: N = 6 - 18. Simulation data show that the ML method performs better than the method based on order statistics, especially in difficult situations, e.g. large expected censored proportion h_ex (h_ex ≥ 50%) and for small sample size (N = 6).

C. Perrin, Y. Vander Heyden, M. Maftouh and D.L. Massart.

"Rapid screening for chiral separations by short-end injection capillary electrophoresis using highly sulfated cyclodextrins as chiral selectors". Electrophoresis, 22, 3203-3215 (2001)

A rapid strategy for the separation of various chiral drugs and drug precursors by capillary electrophoresis using highly sulfated cyclodextrins was developed. Short-end injection was used in order to reduce dramatically analysis time. Highly sulfated cyclodextrins appeared to be very versatile selectors. 62 out of the 67 compounds investigated could be baseline resolved with usually very high selectivity.

E. Smet, L. Staelens, Y. Vander Heyden, W.R.G. Baeyens, H.Y. Aboul-Enein, G. Van Der Weken and A.M. García-Campańa.

"Optimization of the chiral separation of some 2-arylpropionic acids on an avidin column by modeling a combined response". Chirality, 13, 556-567. (2001).

The enantiomeric separation of some nonsteroidal aniinflammatory drugs on an avidin column was investigated. An experimental design approach was used to evaluate the effects of pH, concentration of organic modifier, and buffer concentration, on the analysis time and the resolution, as well as to model these responses. Furthermore, the central composite design results were combined in a multicriteria decision-making approach in order to obtain a set of optimal experimental conditions leading to the most desirable compromise between resolution and analysis time.

F. Estienne, Y. Vander Heyden and D.L. Massart.

"Chemometrics and modeling". Computational Chemistry Column, Chimia, 55, 70-80 (2001)

Modeling is an important part of chemometrics. It consists in relating two or more characteristics in such a way that the obtained model represents reality as well as possible. The classical least squares regression method is described and compared to such methods as orthogonal least squares, inverse and robust regression. Genetic algorithms are presented as a way of selecting variables for multivariate regression. Regression methods such as principal component regression and partial least squares are introduced, as well as the use of N-way principal components.

E. Adams, D. Coomans, J. Smeyers-Verbeke and D.L. Massart.

"Application of Linear Mixed Effects Models to the Evaluation of Dissolution Profiles ". Int. J. Pharmaceutics, 226, 107-126 (2001)

The performance of linear mixed effects models for the comparison of dissolution profiles is investigated. This type of models is frequently used by statisticians, but rather unknown to people that perform dissolution tests. Hence, an extensive theoretical part was added to make the methodology more accessible. After repeated measures ANOVA, the "real" linear mixed effects models are discussed. The theory is applied to two types of dissolution profiles: one corresponding to an immediate and another to a slow release formulation. The standard settings of the statistical software (S-plus) were used as much as possible. Suggestions are given to solve problems encountered during model fitting. It was found that the statistical limits are much more discriminative than the f₂ factor.

J.J.N. Cirunay, Y. Vander Heyden and J. Plaizier-Vercammen.

"LC Separation of Calcipotriol from its Photodegradation Products and Protection Possibilities using Adjuvants". Journal of Pharmaceutical and Biomedical Analysis, 26, 31-42 (2001).

Mobile phase optimization and reversed-phase column characteristics were used to separate photodegradation products from calcipotriol. Separation was obtaincd with an acetonitrile/water (53:47, v/v) mobile phase on a C₁₈ Hypersil ODS column (250 mm, 4.6 mm id, 5 mm) and a flow rate of 1 ml/min. Using this system, the influence of commonly used solvents in dermatology on degradation was studied. The addition of a UV filter (2-hydroxy-4-methoxybenzophenone) was also evaluated for its possible protective effect to light exposure. The sunscreen affords a protection proportional to the filter concentration.

   V. Schoonjans and D.L. Massart.
"Combining Spectroscopic data (MS, IR): Exploratory Chemometric Analysis for Characterizing Similarity/Diversity of Chemical Structures". J. Pharm. Biomed. Anal., 26, 225-239 (2001)

Combined infrared-mass spectra have been used to examine a small data set of synthetic substances in order to elucidate whether a combination of spectral descriptors yield better classification and similarity predictions than their corresponding individual spectral descriptors. To visualize the structure of the data, Principal Component Analysis was applied. Hierarchical upgma-cluster analysis was also used for data classification.

B. Rambali, L. Baert and D.L. Massart.

"Using experimental design to optimize the process parameters in fluidized bed granulation on semi-full scale". Int. J. Pharmaceutics, 220, 149-160 (2001)

A face-centered central composite design was applied in order to optimize the granulation process on a semi-full scale for the geometric mean granule size. The granulation process variables investigated were: inlet air temperature, inlet airflow rate, spray rate and inlet air humidity. Based on the process variables, the theoretical powder bed moisture content after the spraying process and a measure for the droplet size were determined. Multiple regression modeling was used to develop two models for granule size: an empirical model, based on the four process parameters, and a fundamental model, based on the balance between the granule growth affected by the theoretical powder bed moisture content and the droplet size and the breakage effect of the airflow rate.

B. Rambali, L. Baert, E. Jans and D.L. Massart.

"Influence of the roll compactor parameter settings and the compression pressure on the buccal bio-adhesive tablet properties". Int. J. Pharmaceutics, 220, 129-140 (2001)

Miconazole buccal tablets were prepared via a dry ganulation process. By applying a factorial design (2⁴), the roll compactor parameters were optimised for the tablet strength. The compaction force and the roll type significantly affected the tablet strength. Afterwards, a quarter fractional factorial design (2^5-2) was applied, consisting of the four compactor parameters and additionally the compression pressure, in order to optimise these parameters for the dissolution profile and the buccal bio-adhesion characteristics (bio-adhesive force and energy). In order to evaluate the dissolution profiles properly, the similarity factor between sample and a zero-order release reference profile was used.

   V. Pravdova, B. Walczak, D.L. Massart, H. Robberecht, R. Van Cauwenbergh, P. Hendrix and H. Deelstra.
"Three-way principal component analysis for the visualization of trace elemental patterns in vegetables after different cooking procedures". J. Food Composition and Analysis,14, 207-225 (2001)

Studies performed in food chemistry, mainly the nutritional content of foodstuffs, often produce to large and complex data sets. The 3-way data structure (elemental content x type of vegetable x cooking style) is described here. To visualize the main trend in this data 3-way PCA was applied and compared to better known 2-way PCA. It is shown that the vegetables can be classified according to the elemental content and effect of cooking procedure on that content. It is concluded that 3-way PCA can be a useful tool for exploratory data analysis of nutritional tables.

   V. Simeonov, P. Barbieri, B. Walczak, D.L. Massart and S. Tsakovski.
"Environmetric modeling of a potable water data set". Toxicol. & Environm. Chemistry, 79, 55-72 (2001)

The present study deals with the application of different multivariate statistical approaches to a pot water data set from Athens, Greece. The set represents analytical monitoring data for 16 potable water parameters (turbidity, pH, free acidity, total hardness, chloride, sulfate, nitrate, nitrite, calcium, magnesium, sodium, chemical oxygen demand, dry residue, conductivity, phosphate and silicon dioxide) measured in samples collected at 17 different sampling sites during 6 bimonths intervals. Both unsupervised and supervised environmetrical methods were used for data treatment: two- and multi-way principal component analysis (PCA and MPCA-Tucker 3 model) and CART (partitioning tree), respectively. It has heen found that "rural" and "urban" conditional factors are responsible for the site spatial separation and "salinity" and "turbidity" latent factors - for the data structuring. Additionally, a separation in "spring/summer" and "autumn/winter" factors is observed determining the seasonal behavior of the water sources.

S. Lauwers, Y. Vander Heyden and B. Rombaut.

"Screening of an enterovirus specific RT-PCR ELISA method for the quantification of enterovirus genomes in human body fluids by means of a three-level experimental design". J. Pharm. Biom. Analysis, 25, 131-142 (2001).

To obtain a low detection limit for a quantitative enterovirus specific RT-PCR ELISA assay, optimal reaction conditions needed to be determined. This was done by investigating the influence of thirteen factors, selected from RT and PCR, in a multivariate approach by means of a three-level screening design, derived from a three-level Plackett-Burman design. Optimal reaction conditions were determined by calculation and evaluation of the effects of the different factors on the measured absorbance of the ELISA detection.

   M Daszykowski, B. Walczak and D.L. Massart.
"Looking for natural patterns in data.- Part 1. Density-based approach". Chem. Intel. Lab. Syst., 56, 83-92 (2001)

A density-based unsupervised clustering approach for detecting nartural patterns (NP) in data is presented. NP works for arbitrary clusters, is a single-scan technique, requires no presumptions regarding data distribution and requires only one input parameter, which describes the minimal number of objects, considered as cluster. NP can be applied not only for data clustering, but also for the identification of outliers.

B. Rambali, L. Baert, D. Thoné, D.L. Massart.

"Using experimental design to optimise the process parameters in fluidised bed granulation". Drug dev. & ind. pharm., 27(1), 47-55 (2001).

Parameters were screened for a small-scale granulation process for their effect on the yield of granules between 70 and 500 µm and the geometrical granule mean size. First a Plackett-Burman design was applied to screen the inlet air temperature, the inlet flow rate, the spray rate, the nozzle air pressure, the nozzle spray diameter, and the nozzle position. The key process parameters were the inlet flow rate and the spray rate and probably also the inlet air temperature. A fractional factorial design (2^5-2) was applied to screen the remaining parameters plus the nozzle aircap position and the spraying time interval. The fractional factorial design showed that the nozzle air pressure was also important. As the target values for the granule yield and the geometric mean granule size were reached during the screening experiments, further optimization was not considered necessary.

A. Detroyer, Y. Vander Heyden, S. Carda-Broch, M.C. García-Alvarez-Coque and D.L. Massart.

"Quantitative Structure-Retention and Retention-Activity Relationships of ß-Blocking Agents by Micellar Liquid Chromatography" J. Chrom. A, 912, 211-221 (2001)

ß-Blocking agents showing a large range of hydrophobicity (log P) were subjected to micellar liquid chromatography. An investigation towards the correlation between log P and the retention factor extrapolated to a mobile phase free of micelles and organic modifier was conducted. The use of an interpolated retention factor or the retention factor for specific individual experimental mobile phases was however advantageous because in the selected mobile phases the retention factors of all ß-blocking agents were measurable. Good correlations with log P and with in vitro biological parameters were seen.

   Y. Vander Heyden, A. Nijhuis, J. Smeyers-Verbeke, B.G.M. Vandeginste and D.L. Massart.
"Guidance for robustness/ruggedness tests in method validation". J. Pharm. Biom. Anal., 24, 723-753 (2001).

The manuscript gives guidance in setting-up and interpreting a robustness/ruggedness test. The different steps are discussed and illustrated. The recommendations given are based on approaches found in the literature, several case studies performed by the authors and discussions of the authors within a commission of the French SFSTP (Société Française des Sciences et Techniques Pharmaceutiques). A worked-out example is given of a robustness test case study that is set up and interpreted according to the guidelines.

   J.A. Fernández Pierna, D.L. Massart, O.E. de Noord and Ph. Ricoux.
"Direct orthogonalization : some case studies" Chem. Intell. Lab. Syst., 55, 101-108 (2001)

Using a Direct Orthogonalization in data before a PLS model, the total number of components is the same as when PLS alone is used giving a similar RMSECV. The results do not change or become worse when PCR is used. However, with this pre-processing method, a better understanding of the variation present in the data is obtained.

M. Praisler, I. Dirinck, J. Van Bocxlaer, A. De Leenheer and D.L. Massart.

"Computer-aided screening for hallucinogen and stimulant amphetamines with gas chromatography - Fourier Transform Infrared Spectroscopy (GC-FTIR)", J. Anal. Toxicol., Vol. 25, 45-56 (2001)

An expert system applied as a screening test for amphetamine analogues found in recreational-drug exhihits (tablets or powders) is described. The knowledge base defining the reference Fourier transform infrared spectroscopic (FTIR) spectral patterns has been built according to criteria encompassing toxicological, pharmacological, and neurochemical aspects. The class identity of a compound is determined within seconds using soft independent modeling of class analogy (SIMCA). The specificity and the selectivity of the screening test, evaluated by testing 159 toxicologically relevant compounds, are discussed.

   E. Adams, R. De Maesschalck, B. De Spiegeleer, Y. Vander Heyden, J. Smeyers-Verbeke and D.L. Massart.
"Evaluation of Dissolution Profiles using Principal Component Analysis", Int. J. of Pharmaceutics, 212, 41-53 (2001)

The performance of principal component analysis for the evaluation of dissolution profiles is examined and compared to other methods such as the similarity factor and the calculation of the area under the curve. Both simulated and real data from the pharmaceutical industry are used. The PCA scores plots provide information about the between- and within-batch variations of the dissolution curves. Differences in level can be observed in PC1 and differences in shape in PC2. Irrelevant irregularities, which have a strong influence on the similarity factor, are neglected in PC1/PC2. To detect outliers in a set of dissolution curves, PCA was preferred above the Hotelling's T2 test. In general, PCA is found to be a useful technique to examine dissolution data visually, but however, it does not contain criteria to decide if batches are similar or not. This can be done by combining PCA with the resampling with replacement or bootstrap method to construct confidence limits.

C.A. De Vriendt, D.K. Van Sassenbrouck, M.T. Rosseel, E.J. Van De Velde, A.G. Verstraete, Y. Vander Heyden and F.M. Belpaire.

"Development and validation of an HPLC method for the determination of gamma-hydroxybutyric acid (GHB) in rat plasma", Journal of Chromatography Biomed Applications, 752, 85-90 (2001).

A method for the determination of gamma-hydroxybutyric acid (GHB) in rat plasma was developed. GHB was isolated from plasma using anion-exchange SPE columns. The chromatographic separation was performed on a C₁₈ Aqua column. The lower limit of quantification was 10 µg/ml using 60 µl of plasma. The linearity of the calibration curves was satisfactory. The within-day and between-day precision were <10% (n=24), the accuracy was nearly 101%. Plasma concentrations in rats after GHB infusion determined by HPLC - UV were compared with those of a validated gas chromatographic - mass spectrometric method by orthogonal distance regression. A good correlation was observed and no significant differences from 0 and 1 for the intercept and slope, respectively.

   V. Schoonjans, F. Questier, Q. Guo, Y. Vander Heyden and D.L. Massart.
"Assessing Molecular Similarity/Diversity of Chemical Structures by FT-IR Spectroscopy", J. Pharm. Biom. Analysis, 24, 613-627 (2001)

FT-IR spectra have been investigated for their ability to distinguish compounds which are chemically diverse and to produce clusters of compounds which makes sense chemically. Principal Component Analysis was applied to the analysis of a small database of FT-IR spectra, to visualize the structure of the data. Sequential Projection Pursuit was applied to detect outlying objects in the data. Finally, cluster analysis of these spectra, depending on unweighted pair-group average linkage, was carried out.

   B. Massart, Q. Guo, F. Questier, D.L. Massart, C.Boucon, S. de Jong and B.G.M. Vandeginste.
"Data structures and data transformations for clustering chemical data", TrAC, vol. 20 (1), 35-41 (2001)

The quality of a clustering of chemical data is determined by a proper choice of distance measures and data transformations. The latter aspect is often neglected. It is shown that the V-shaped data structure that is often obtained in a principal component analysis of chemical data may indicate that the clustering of the raw data can lead to classifications that are not relevant from a chemical point of view and that the log double centering transform should be considered as a possible alternative.

M.I. Griep, T.F. Mets, K. Collys, I. Ponjaert-Kristoffersen and D.L. Massart.

"Risk of malnutrition in retirement homes elderly persons measured by the "Mini-Nutritional Assessment".", J. of Gerontology: Medical Sciences, 55A, M57-M63 (2000).

   F. Estienne, D.L. Massart, N. Zanier-Szydlowski and P. Marteau.
"Multivariate calibration with Raman spectroscopic data : a case study", Analytica Chimica Acta, 424, 185-201 (2000).

The general aim of this study was to improve the precision in the monitoring of an industrial process separating p-xylene from mainly other C₈ aromatic compounds. The concentrations of six constituents being currently evaluated with a classical calibration method, inverse calibration linear methods were applied on a synthetic data set, in order to evaluate the improvement in prediction such methods could yield. The methods tested included among others Principal Component Regression with Variable Selection, Partial Least Square Regression or Multiple Linear Regression with variable selection (Stepwise or based on Genetic Algorithm). Methods based on selected wavelengths happened to be very efficient as the obtained models can be expected to be very robust toward experimental conditions. However, issues related to random correlation arose due to the important noise in the data. Several ways of solving this problem were investigated, including the use of signal pre-processing (smoothing and filtering in the Fourier or Wavelets domain), and the use of an improved variable selection algorithm based on the selection of spectral windows instead of single wavelengths when this leads to a better model. Stepwise variable selection applied to spectra denoised in the Fourier domain gave the best results.

   A. Detroyer, V.Schoonjans, F. Questier, Y. Vander Heyden, A.P. Borosy, Q. Guo and D.L. Massart.
"Exploratory chemometric analysis of the classification of pharmaceutical substances based on chromatographic data", J. Chromatography A, 897, 23-36 (2000)

A chemometric study has been conducted on a published data set. The set consists of the retention times of 83 substances, from 5 pharmacological families, on 8 HPLC systems. Principal Component Analysis, Clustering and Sequential Projection Pursuit were applied. Thus the extent, to which the combination of chromatography and chemometrics allows to make conclusions about pharmacological activities of (candidate) drugs and the contribution of the different HPLC systems considered, was investigated.

   E. Hund, D.L. Massart and J. Smeyers-Verbeke.
"Inter-laboratory studies in analytical chemistry", Analytica Chimica Acta, 423, 145-465 (2000)

The paper reviews different types of interlaboratory studies: method-performance studies (collaborative trials), laboratory performance studies (proficiency tests), collaborative bias evaluation, inter-laboratory evaluation of to-be standard methods and certification studies for reference materials. Besides the classical evaluation methods using outlier-tests, robust statistics and graphical methods are considered.

M. Praisler, I. Dirinck, J. Van Bocxlaer, A. De Leenheer and D.L. Massart

"Pattern recognition techniques screening for drugs of abuse with gas chromatography - Fourier transform infrared spectroscopy" Talanta, 53, 177-193 (2000)

M. Praisler, I. Dirinck, J. Van Bocxlaer, A. De Leenheer and D.L. Massart

"Identification of novel illicit amphetamines from vapour-phase FTIR spectra. A chemometrical solution" Talanta, 53, 155-170 (2000)

Y. Vander Heyden, C. Perrin and D.L. Massart.

"Optimization strategies for HPLC and CZE". Chapter 6 in "Handbook of Analytical Separations: Separation Methods in Drug Synthesis and Purification" (Ed. Valkó K.), Vol 1, 1-50 (2000)

This chapter describes experimental strategies that can be followed during the screening, optimisation and robustness testing phases of method development. The responses considered are discussed, as well as univariate and multivariate optimisation strategies (screening designs, full factorials, response surface designs, mixture designs, and simplex sequential approach) and the applied models.

V. Simeonov, D.L. Massart, G. Andreev and S. Tsakovski.

"Assessment of metal pollution based on multivariate statistical modeling of 'hot spot' sediments from the Black Sea". Chemosphere, 41, 1411-1417 (2000)

The paper deals with application of different statistical methods like cluster and principal components analysis (PCA), partial least squares (PLSs) modeling. These approaches are an efficient tool in achieving better understanding about the contamination of two gulf regions in Black Sea. As objects of the study, a collection of marine sediment samples from Varna and Bourgas "hot spots" gulf areas are used. The use of cluster and PCA make it possible to separate three zones of the marine environment with different levels of pollution. The extraction of four latent factors offers an interpretation of the pollution sources and separates natural from anthropogenic factors. Finally, the PLSs modeling gives a better opportunity in predicting contaminant concentration on tracer (or tracers) element as compared to the one- dimensional approach of the baseline models.

   Y. Vander Heyden, S. Kuttatharmmakul, J. Smeyers-Verbeke and D.L. Massart.
"Supersaturated designs for robustness testing". Anal. Chem.,72, Nr 13, 2869-2874 (2000)

Supersaturated designs examine a number of factors that exceed the number of experiments. Even the main effects are confounded, when estimated. However the total variance of a response can in principle be used as a measure for the robustness of a method. It was examined if the variance estimated from a supersaturated design is similar to the one from a Plackett-Burman design. This was found to be the case, which means that the estimated variance describes well the variation in the response caused by the variation in the factors.

   C. Perrin, M.G. Vargas, Y. Vander Heyden, M. Maftouh and D.L. Massart.
"Fast development of separation methods for the chiral analysis of amino acid derivatives using capillary electrophoresis and experimental design". J. Chromatogr. A, 883, 249-265 (2000)

The enantiomeric separation of a series of amino acid derivatives was investigated using an experimental design strategy. Three modified cyclodextrins (CDs) were used as chiral selectors: hydroxypropyl-β-CD, carboxymethyl-β-CD and sulfobutylether-β-CD. The type of CD, its concentration, the pH of the electrolyte and its content in methanol were studied in the design. A general scheme for the chiral separation of various drugs or precursors is proposed.

   S. Kuttatharmmakul, D.L. Massart and J. Smeyers-Verbeke.
"Comparison of alternative measurement methods : determination of the minimal number of measurements required for the evaluation of bias by means of interval hypothesis testing". Chem. Intell. Lab. Systems, 52, No. 1, 61-73 (2000)

The classical way of bias evaluation by means of point hypothesis testing (i.e. the significance of the bias is tested against a single fixed point, namely zero) has a major drawback that the risk beta, of wrongly adopting a method which in fact has an unacceptable bias, can be unacceptably large. To solve this drawback, the interval hypothesis testing, which the significance of bias is tested against the interval of acceptance values, has been proposed. However, the application of the interval hypothesis testing may lead to the wrong rejection of a method that in fact has an acceptable bias. To control this risk of wrong rejection, an appropriate sample size must be applied. This article describes the procedures to determine the sample size required for the evaluation of bias from the comparison of alternative measurement methods by means of interval hypothesis testing. Formulae for the sample size determination are provided and evaluated by means of simulations.

   K. De Braekeleer, J.R. Torres-Lapasío and D.L. Massart.
"Improved purity assessment of HPLC-DAD data for overcoming the presence of the non-linearity artefact". Chemom. Intell. Lab. Syst., 52, No.1, 45-59 (2000)

Orthogonal Projection (OPA) and Needle Search (NS) have been applied to assess the purity of peaks in a spectrochromatogram recorded with a high-performance liquid chromatograph coupled with a diode array detector. The introduction of an anomalous number of compounds in the peak purity assays points out to the existance of an non-linearity artefact in the DAD data set, which leaves unexplained contributions greater than the contributions associated to the known impurities. Decision criteria and certain modifications were introduced in OPA and NS in order to distinguish the artefact peaks from the peaks corresponding to minor components.

   C. Vannecke, A. Nguyen Minh Nguyet, M.S. Bloomfield, A.J. Staple, Y. Vander Heyden & D.L. Massart.
"Development and optimisation of a flow injection assay for fluticasone propionate using an asymmetrical design and the variable-size simplex algorithm". J. Pharm. Biomed. Anal.,23, 291-306 (2000)

A flow injection analysis method is described to determine fluticasone propionate (a thio-ester). The method is based upon the reaction of o-phthalaldehyde with a primary amine and a thiol, the latter formed after a basic hydrolysis of the thio-ester. First a screening is executed to select the significant factors and in a second step these factors are optimised with the variable-size simplex algorithm. In the screening step, a two-level fractional factorial design is compared with an asymmetrical design containing the same number of experiments, but in which one factor is at three levels. Complete optimisation was carried out for both UV and fluorescence detection.

   Q. Guo, W. Wu, F. Questier, D.L. Massart, C. Boucon and S. de Jong.
"Sequential Projection Pursuit using Genetic Algorithms for Data Mining of Analytical Data". Anal. Chem., 72, 2846-2855 <2000).

Sequential projection pursuit (SPP) is proposed to detect inhomogeneities (clusters) in high-dimensional analytical data. Such inhomogeneities indicate that there are groups of objects (samples) with different chemical characteristics. The method is compared with principal component analysis (PCA). In SPP, latent variables are obtained sequentially according to their importance measured by the entropy index. This involves an optimization step, which is achieved by using a genetic algorithm. The performance of the method is demonstrated on near-infrared and gas chromatography data sets.

W. Wu & R. Manne.

"Fast regression methods in a Lanczos (or PLS-1) basis. Theory and applications". Chem. Intell. Lab. Syst., 51, 145-161 (2000)

In order to improve the calibration speed for very large data sets, novel algorithms for principal component regression (PCR) and partial-least-squares (PLS) regression are presented. They use the Lanczos or PLS-1 transformation to reduce the data matrix X to a small bidiagonal matrix (R), after which the small tridiagonal matrix (R'R) is diagonalized and inverted. The complexity of the PCR model may be optimized by cross-validation (PCRL) but also using simpler and faster recipes based upon round-off monitoring and model fit (PCRF). A similar fast PLS procedure (PLSF) is also presented. Calculations are made for near infrared spectroscopy (NIR) data sets and compared with PCR with feature selection (PCRS) based on correlation and with de Jong's simple partial least squares (SIMPLS). The Lanczos-based methods are considerably faster.

S. Toasaksiri, D.L. Massart and Y. Vander Heyden.

"Study of method validation criteria in a capillary electrophoresis method for the separation of non-steroidal anti-inflammatory drugs". Analytica Chimica Acta, 416, 29-42 (2000).

The separation of three non-steroidal anti-inflammatory drugs (NSAIDs) by a capillary electrophoresis (CE) method was examined and validated. The method was first optimised using experimental designs (half-fraction factorial designs for five factors). The validation of the method showed good results for the selectivity, linearity, system repeatability and bias. The calibration lines from peak areas and corrected peak areas were linear while those from peak heights curved. In some concentration ranges the data were heteroscedastic. The ordinary least squares (OLS) and weighted least squares (WLS) regressions were used and compared. Half-fraction factorial designs were also applied to test the robustness of the method.

   K. De Braekeleer, A. de Juan, F. Cuesta Sánchez, P.A. Hailey, D.C.A. Sharp, P. Dunn and D.L. Massart.
"Determination of the end point of a chemical synthesis process using on-line measured mid-infrared spectra". Appl. Spectrometry 54 (4), 601-607 (2000)

An organic synthetic process was monitored on-line by mid-infrared spectroscopy. The Orthogonal Projection Approach (OPA) and Principal Component Analysis (PCA) were used to detect time regions in the synthesis process where the reaction reached a steady state (constant temperature and concentration of compounds). Multivariate Curve Resolution Alternating Least Squares (MCR-ALS) was applied on the steady state regions to obtain information about the chemical composition of the reaction medium.

   V. Centner, G. Verdú-Andrés, B. Walczak, D. Jouan-Rimbaud, F. Despagne, L. Pasti, R. Poppi, D.L. Massart and O.E. de Noord.
"Comparison of multivariate calibration techniques applied to experimental NIR data sets". Appl. Spectrometry 54 (4), 608-623 (2000).

B. Walczak and D.L. Massart.

"Calibration in wavelet domain" Chapter 15 in "Wavelets in chemistry" (Ed. Walczak) 323-349 (2000)

B. Walczak and D.L. Massart.

"Joint basis and joint best-basis for data sets" Chapter 7 in "Wavelets in chemistry" (Ed. Walczak) 165-176 (2000)

M. Baret, D.L. Massart, P. Fabry, F. Conesa, C. Eichner and C. Menardo.

"Application of neural network calibrations to an halide ISE array". Talanta, 51, 863 - 877 (2000)

The Neural Network (NN) technique was applied to the calibration of an ion selective electrode (ISE) array comprising a bromide selective electrode, two chloride ISEs and one thiocyanate ISE. The samples were synthetic mixture solutions. The NN method allowed to perform the calibration without estimating the coefficients of the Nikolskii - Eisenman theoretical relation. The results obtained using this method were better than those obtained using linear multivariate calibration methods.

   E. Hund, Y. Vander Heyden, M. Haustein, D.L. Massart & J. Smeyers-Verbeke.
"Robustness testing of an HPLC assay : comparison of fractional and asymmetrical factorial designs". J. Chromatography A, 874, 167-185 (2000)

Robustness tests were performed on a RP-HPLC assay in order to compare different experimental designs. Two two-level fractional factorial designs (different resolutions) were used to study the influence of procedure-related factors. Using asymmetrical factorial designs, the factors chromatographic column manufacturer (4 levels) and instrument (3 levels) were stepwise included in the study. The significance of the factor effects was determined using two types of error estimates, and by means of half-normal plots. The asymmetrical designs turned out to be an efficient and economic method to examine the influence of factors at different numbers of levels in the robustness testing of analytical methods.

   A. Garrido Frenich, M. Martínez Galera, J.L. Martínez Vidal, D.L. Massart, J.R. Torres-Lapasío, K. De Braekeleer, Ji-Hong Wang and P.K. Hopke.
"Resolution of multicomponent peaks by orthogonal projection approach, positive matrix factorization and alternating least squares". Anal. Chim. Acta, 411, 145-155 (2000).

The application of orthogonal projection approach (OPA), alternating least squares (ALS), and positive matrix factorization (PMF) to resolve HPLC-DAD data into individual concentration profiles and spectra is discussed. OPA, ALS and PMF have been applied using a single matrix (two-way data) or an augmented matrix containing several data matrices simultaneously. The results obtained with the different resolution methods are compared and evaluated.

   F. Despagne, P. Chabot and D.L. Massart.
"Development of a robust calibration model for nonlinear in-line process data". Anal. Chem., 72, 1657-1665 (2000).

A study involving partial least squares, locally weighted regession, and neural networks has been performed in order to implement a calibration model on an industrial process. The models were designed to predict the water content in a reactor during a distillastion process, using in-line measurements from a near-infrared analyzer. Curved effects due to changes in temperature and variations between the different batches make the problem particular callenging. The use of a monitoring set with neural networks does not always prevent overfitting. Therefore, we developed a model selection criterion based on the determination of the median of monitoring error over replicate trials. The back-propagation neural network models selected were found to outperform the other methods.

   S. Kuttatharmmakul, J. Smeyers-Verbeke, D.L. Massart, D. Coomans & S. Noack.
"The mean and standard deviation of data, some of which are below the detection limit : an introduction to maximum likelihood estimation". TrAC, 19, no. 4, 215-222 (2000)

In this article, the principle of maximum likelihood estimation (MLE) is introduced. It is illustrated by an application of the maximum likelihood method for the estimation of the mean and standard deviation of a singly censored data set. An example of the determination of an impurity in a raw material, where the measurements are carried out around the detection limit and some of these fall below it, is given to illustrate the application of MLE to a singly censored data set.

   A. Candolfi & D.L. Massart.
"Model updating for the identification of NIR-spectra from a pharmaceutical excipient". Appl. Spectroscopy, 54 (1), 48-53 (2000).

The effect of model updating on the identification of a pharmaceutical excipient based on its near-infrared (NIR) spectra has been investigated. A pragmatic updating approach, consisting of adding stepwise newly available samples to the training set and rebuilding the classification model, was applied. Its performance is compared for three pattern recognition methods: the wavelength distance method, the Mahalanobis distance method and the SIMCA (soft independent modeling of class analogy) residual variance method.

   Y. Vander Heyden, P. Vankeerberghen, M. Novic, J. Zupan & D.L. Massart.
"The application of Kohonen neural networks to diagnose calibration problems in atomic absorption spectrometry". Talanta, 51, 455-466 (2000)

Calibration lines measured in atomic absorption spectrometry are not always acceptable. They can contain outliers, have a bad precision or can be curved. In this paper Kohonen neural networks were used to classify the calibration lines. The results were compared with the classification made by a decision support system based on classical statistical methods. The prediction capabilities for both approaches, relative to a visual classification, were found to be comparable with an error rate of <10%.

   J.A. Fernández Pierna & D.L. Massart.
"Improved algorithm for clustering tendency". Anal. Chim. Acta, 408, 13-20 (2000)

A modified Hopkins algorithm for clustering tendency is described. With this modified algorithm, the automatic detection of clusters, before multivariate calibration, is carried out.

   F. Despagne, D.L. Massart, M. Jansen and H. van Daalen.
"Intersite transfer of industrial calibration models". Anal. Chim. Acta, 406, 233-245 (2000).

Two instrument standardisation methods, the piecewise direct standardisation and a method based on neural networks, are compared for the transfer of near-infrared spectra between two sites. Some important issues that can affect transfer, such as representativity of standardisation samples are discussed in detail. It is shown that both methods lead to different reconstruction results in the presence of structured background noise.

M.I. Griep, T.F. Mets and D.L. Massart.

"Effects of flavour amplification of Quorn® and yoghurt on food preference and consumption in relation to age, BMI and odour perception". Brit. J. Nutrition, 83, 105-113 (2000).

Since the food habits of many elderly people are inadequate, it was tested whether flavour amplification induces changes in preferences for and consumption of food and thus might result in a nutritional benefit. Two panels: one of 260 and one of 120 suhjects, aged 19 - 98 years, took part in the study in which preferences for flavour-amplified yoghurt and Quorn^® were measured. The percentage of subjects preferring the high flavour levels increased with age. It is concluded that flavour amplification of food for older adults deserves attention, but specific approaches, which are tailored to the candidate food systems and older adult target groups, are needed.

   J.R. Torres-Lapasió, D.L. Massart, J.J. Baeza-Baeza and M.C. García-Alvarez-Coque.
"A three-factor optimisation strategy for micellar liquid chromatography". Chromatographia, 51 No 1/2, 101-110 (2000).

An interpretive optimisation methodology for micellar liquid chromatography (MLC) is shown, taking into account pH, surfactant (sodium dodecyl sulphate) and organic modifier (propanol) concentrations. A strategy is proposed which has the advantage of allowing partial optimisations in two factors at constant pH levels, before completing the three-factor experimental design. The sequential strategy adapts retention model and design to the difficulty of the separation problem, which saves further experimental work if any of the developed partial optimisations succeeds before developing the three-factor design. In a favourable case, four experiments can be enough.

   V. Schoonjans, F. Questier, A.P. Borosy, B. Walczak, D.L. Massart & B.D. Hudson.
"Use of mass spectrometry for assessing similarity/diversity of natural products with unknown chemical structures". J. Pharmac. Biomed. Anal., 21, 1197-1214 (2000).

An evaluation whether clustering techniques, using mass spectral data, can be applied for assessing similarity/diversity of chemical compounds was performed and we investigated how much information is lost by using these analytical characteristics instead of the chemical structure for characterizing similarity/diversity. PCA and Ward's hierarchical method were used to cluster a small data set of synthetic substances.

M. Praisler, I. Dirinck, J. Van Bocxlaer, A. De Leenheer & D.L. Massart.

"Exploratory analysis for the automated identification of amphetamines from vapour-phase FTIR spectra". Anal. Chim. Acta., 404, 303-317 (2000).

An exploratory analysis was performed in order to evaluate the feasibility of building an expert system automating the identification of the basic skeleton of amphetamines necessary in the investigation of drugs of abuse. Emphasis was placed on the selection of the sample preparation and data processing techniques yielding the best differentiation and recognition of the structural patterns of amphetamines.

   R. De Maesschalck, D. Jouan-Rimbaud & D.L. Massart.
"Tutorial - The Mahalanobis distance". Chemom. Intell. Lab. Syst., 50, 1-18 (2000).

The Mahalanobis distance (MD), in the original and principal component (PC) space, are examined and interpreted in relation with the Euclidean distance (ED). Techniques based on the MD and applied in different fields of chemometrics such as in multivariate calibration, pattern recognition and process control are explained and discussed.

   E. Hund, Y. Vander Heyden, M. Haustein, D.L. Massart & J. Smeyers-Verbeke.
"Comparison of several criteria to decide on the significance of effects in a robustness test with an asymmetrical factorial design". Anal. Chim. Acta., 404, 257-271 (2000).

A robustness test was performed for a reversed-phase HPLC method. It is based on an asymmetrical factorial design that allows to examine simultaneously a factor (column manufacturer) at four levels and twelve factors at two levels. Different statistical methods to decide on statistical significance of the observed effects are compared with a graphical evaluation by means of the half-normal plot.

V. Simeonov, S. Tsakovski & D.L. Massart.

"Multivariate statistical modeling of coastal sediments data". Toxicological & Environmental Chemistry, 72, 81-92, (1999).

A coastal sediments data set (127 x 16 elements) was treated by the use of multivariate statistical approaches like PCA and PLS in order to obtain information about possible latent factors influencing the process of sediment formation. Three main latent factors seem to regulate the sediment formation in the region of interest - a "natural", a "biogenic", an "anthropogenic". An assumption for the fourth latent factor (" organic anthropogenic") could be made.
The PLS modeling of the data set indicates that adequate models could be obtained to predict the amount of polluting substances in the sedimentation by the use of four basic tracers (iron, aluminium, grain size and total organic carbon) as independent variables.

M. Baret, D.L. Massart, P. Fabry, C. Menardo & F. Conesa.

"Halide ion-selective electrode array calibration". Talanta, 50, 541-558 (1999)

The calibration of several ions measured with an ion selective electrodes (ISE) array has been carried out in the presence of interferents using an experimental design and multivariate calibration methods.

   R. De Maesschalck, F. Estienne, J. Verdú-Andrés, A. Candolfi, V. Centner, F. Despagne, D. Jouan-Rimbaud, B. Walczak, D.L. Massart, S. de Jong, O.E. de Noord, C. Puel & B.M.G. Vandeginste.
"The development of calibration models for spectroscopic data using principal component regression". Internet Journal of Chemistry, 2, 19 (1999). URL:"http://www.ijc.com/articles/1999v2/19/"

This article aims at explaining how to develop a calibration model for spectroscopic data analysis by Principal Component Regression (PCR). The principle of PCR is to perform a Principal Component Analysis (PCA) and then make a Multiple Linear Regression (MLR) on the scores obtained. Before applying the method, the data has to be investigated in oder to detect for instance outliers, clustering tendency or nonlinearities. Once developed, the PCR model can also be optimized using UVE (uninformative variable elimination). How to handle replicates and perform different data preprocessings and/or pretreatments is also explained in this tutorial.

A.I. Rollier, L. Vauterin, P. De Vos, D.L. Massart, L.A. Devriese, L. De Zutter & J. Van Hoof.

"A numerical taxonomic study of the Pseudomonas flora isolated from poultry meat". J. Appl. Microbiol., 87 (1), 15-28 (1999).

Pseudomonas strains were isolated from broiler skin. Phenotypically-based numerical taxonomic techniques were used to characterize the isolates and 36 reference strains. For this purpose, a number of biochemical tests were used. Jaccard clustering revealed the predominance of four major Pseudomonas groups: Ps. fragi, Ps. lundensis, strains belonging to Ps. fluorescens biovars and an unidentified group of strains displaying a high degree of similarity to Ps. fluorescens biovars.

   C. Vannecke, M.S. Bloomfield, Y. Vander Heyden & D.L. Massart.
"Use of experimetal design to optimise a flow injection analysis assay for L-N-monomethylarginine". J. Pharm. Biomed. Anal., 21, 241-255 (1999).

The determination of L-N-monomethylarginine with flow injection analysis was optimised using experimental design. Two different approaches were followed. In the first approach, the chemical reaction was optimised off-line and the optimal chemical conditions were transferred to the FIA-system . In the second approach the reaction and the FIA-system parameters were optimised together. In both approaches critical factors were first identified in a screening design and the simplex algorithm was used for further optimisation.

   P. Barbieri, C.A. Andersson, D.L. Massart, S. Predonzani, G. Adami & E. Reisenhofer.
"Modeling bio-geochemical interactions in the surface waters of the Gulf of Trieste by three-way principal component analysis (PCA). Anal. Chim. Acta, 398, 227-235 (1999).

Data of temperature, salinity, dissolved oxygen, nutrients and chlorophyll measured on samples of surface seawater and collected monthly during 2 years in different sites of the Gulf of Trieste are modeled by means of three-way principal component analysis (PCA). A missing data treatment is described. Physicochemical parameters are described by three different components that explain the effect of the river input of temperature and of metabolic-catabolic activity of the phytoplankton respectively. One spatial component accounts for the gradient of influence of the estuarine waters and three temporal components characterize the main seasonal conditions.

   Y. Vander Heyden, K. De Braekeleer, Y. Zhu, E. Roets, J. Hoogmartens, J. De Beer & D.L. Massart.
"Nested designs in ruggedness testing". J. Pharm. Biomed. Anal., 20, 875-887 (1999).

Nested designs were executed on a high performance liquid chromatography assay to determine tetracycline and related substances in bulk samples of tetracycline. The Nested designs were performed in order to execute a ruggedness test in which mainly non-procedure related factors such as different laboratories, analysts, instruments, columns, days and batches are examined. An alternative interpretation method for nested designs, based on pooled variances, was proposed and found to be appropiate to use for ruggedness testing purposes.

   M.G. Vargas, Y. Vander Heyden, M. Maftouh and D.L. Massart;
"Rapid development of the enantiomeric separation of β-blockers by capillary electrophoresis using an experimental design approach." J. Chromatogr. A, 855 681-693 (1999).

A rapid method for determining the separation conditions for chiral resolution of eleven β-blocking drug substances by capillary electrophoresis is described. Several modified cyclodextrins were applied as chiral selectors: sulfobutyl ether β-cyclodextrin, dimethyl β-cyclodextrin, carboxymethyl β-cyclodextrin and hydroxypropyl β-cyclodextrin. Two different fractional factorial experimental designs were applied: (1) a design examining four factors at three levels (3^4-2) and (2) one examining three factors at two levels (2^3-1). The factors studied were: type of cyciodextrin, cyclodextrin concentration, pH of the background electrolyte and percentage of organic modifier.

   A. Garrido Frenich, J.R. Torres-Lapasío, K. De Braekeleer, D.L. Massart, J.L. Martínez Vidal and M. Martínez Galera.
"Application of several modified peak purity assays to real complex multicomponent mixtures by HPLC-DAD". J. Chromatogr. A, 855, 487-499 (1999)

The detection of multiple closely eluting compounds in a HPLC analysis of a complex multicomponent system was performed by the application of the simple-to-use interactive self-modelling mixture analysis approach (SIMPLISMA), the orthogonal projection approach (OPA) and Needle Search (NS). None of these three approaches succeeded completely when they were applied on the whole data matrix. The OPA and NS methods were slightly modified and applied on smaller parts of the data set. In this way the local complexity of the unresolved mixtures become more favourable which increases the sensitivity of both methods.

   A. Candolfi, R. De Maesschalck, D. Jouan-Rimbaud, P.A. Hailey and D.L. Massart.
"The influence of data pre-procesing in pattern recognition of excipients NIR spectra". J. Pharm. Biomed. Anal., 21, 115-132 (1999)

The effect of data pre-processing (no pre-processing, offset correction, de-trending, standard normal variate transformation (SNV), SNV + de-trending, multiplicative scatter correction, first and second derivative transformation after smoothing) on the identification of ten pharmaceutical excipients is investigated. Four pattern recognition methods are tested in the study, namely the Mahalanobis distance method, the SlMCA residual variance method, the wavelength distance method and a method based on triangular potential functions. The performance of the methods is evaluated on the basis of two NIR data sets.

   E. Hund, D.L. Massart and J. Smeyers-Verbeke.
"Evaluation of the H-point standard additions method (HPSAM) and the generalized H-point standard additions method (GHPSAM) for the UV-analysis for two-component mixtures". J. Pharm. Biomed. Anal., 21 23-42 (1999)

The H-point standard additions method (HPSAM) and two versions of the generalized H-point standard additions method (GHPSAM) are evaluated for the UV-analysis of binary mixtures. Synthetic mixtures (caffeine+phenazone or atovaquone+proguanil) as well as pharmaceutical formulations that contain these substances as active drug substances were used. Both the difficulties related to the methods and the conditions by which acceptable results can be obtained are shown.

   S. Kuttatharmmakul, D.L. Massart and J. Smeyers-Verbeke.
"Comparison of alternative measurement methods". Anal. Chim. Acta, 391, 203-225 (1999)

An approach for the comparison between an alternative measurement method and a reference method has been proposed. Precision and trueness (expressed in terms of bias) are two fundamental performance parameters which are considered. The methodology is based on ISO 5725 part 6 but adapted to the intra-laboratory situation i.e., instead of the reproducibility condition (considered in ISO), the comparison is performed including the (operator + instrument + time)-different intermediate precision and/or the time-different intermediate precision conditions. The proposed approach includes i) the approximation of the optimal sample size required for the comparison experiment, ii) the calculation of different precision estimates and the comparison of these precision estimates between the two methods and iii) the comparison of the trueness between the two methods, in which both applications of the classical point hypothesis testing and the interval hypothesis testing are discussed. Two examples are used to illustrate the proposed approach.

   L. Pasti, B. Walczak, D.L. Massart and P. Reschiglian.
"Optimization of signal denoising in wavelet transform (DWT)". Chem. Intell. Lab. Systems, 48, 21-34 (1999)

A method to optimize the parameters used in signal denoising in the wavelet domain is presented. The method, which is based on cross-validation (CV) procedure, permits to select the best decomposition level and the best wavelet filter function to denoise a signal in the discrete wavelet domain. The proposed method was applied to experimental gravitation field flow fractionation records, and the results were compared with classical low pass filtering in the Fourier domain.

   Y. Vander Heyden, M. Jimidar, E. Hund, N. Niemeijer, R. Peeters, J. Smeyers-Verbeke, D.L. Massart and J. Hoogmartens.
"Determination of system suitability limits with a robustness test". J. Chromatogr. A, 845, 145-154, (1999).

A robustness test was performed on a chromatographic method to identify and assay an active substance and two related compounds in film-coated tablet simulations. A proposal is made to define or re-define system suitability limits based on the results of the robustness test. From the effects found in the robustness test, the experimental conditions yielding the worst result that still is probable to occur when a method is transferred to, for instance, another laboratory, are predicted and the system suitability limits are defined from replicated experiments at these conditions.

   J. Verdú-Andrés, D.L. Massart, C. Menardo and C. Sterna.
"Correction of non-linearities in spectroscopic multivariate calibration by using transformed original variables, Part II : application to principal component regression". Anal. Chim. Acta.,389,115-130 (1999)

The addition of non-linear transformations of the original variables to the original data set is proposed to model strong non-linearities with PCR. This yields always better results than the ones obtained by using only the original variables. In some cases predictive ability increases, in other cases the same predictive ability is achieved with however smaller complexity, giving more parsimonious and robust models.

   D. Jouan-Rimbaud, E. Bouveresse, D.L. Massart and O.E. de Noord.
"Detection of prediction outliers and inliers in multivariate calibration". Anal. Chim. Acta, 388, 283-301 (1999)

Before prediction, one should make sure that the new samples are similar to the calibration samples. Prediction outliers can be of two types, namely samples that are outside of the calibration space, and samples which are situated in a low- (or empty-) density region of the calibration space. The latter type, also called inliers, are not detected by existing methods. In this paper, potential functions are proposed for the detection of inliers and outliers in prediction.

   P. Barbieri, G. Adami, S. Predonzani, E. Reisenhofer and D.L. Massart.
"Survey of environmental complex systems : pattern recognition of physicochemical data describing coastal water quality on the Gulf of Trieste". J. Environmental Monitoring, 1, 69-74 (1999)

   A. Candolfi, R. De Maesschalck, D.L. Massart, P.A. Hailey and A.C.E. Harrington.
"Identification of pharmaceutical excipients using NIR spectroscopy and SIMCA". J. Pharmac. Biomed. Anal., 19, 923-935 (1999).

SIMCA is applied to identify NIR spectra of excipients used in the industry. The performance of SIMCA is described for a situation where the data are not homogeneous because the excipients belong to different batches and come from different suppliers.

W. Prus, Y. Vander Heyden, P. Vankeerberghen, P. Kus, D.L. Massart and T. Kowalska.

"Modelling of solute retention in the RP-HPLC system with the chemically bonded 3-cyanopropyl stationary phase". Chem. Intell. Lab. Syst., 47, 253-266 (1999).

Three models for solute retention are compared (models by Snyder, Schoenmakers and Kowalska) and the retention parameters are correlated with the log P values. Schoenmakers' model performs best in this case. Large deviations from a linear relationship between log k' and log P are obtained in certain circumstances. The deviation is largest when adsorption becomes more important compared to partition in the overall retention.

D.L. Massart.

"Biologische rekenmethoden en hun toepassing in de farmaceutische wetenschappen". Verhandelingen v.d. Koninklijke Academie voor Geneeskunde van België, LX (nr 6), 535-565 (1998).

This Flemish article constitutes an effort to show to pharmacists and medical doctors that biological computation methods (genetic algorithms, neural nets, numerical taxonomy) can be used in drug discovery and pharmaceutical technology.

   B. Walczak and D.L. Massart.
"Rough sets theory". Chem. Intell. Lab. Syst., 47, 1-16 (1999).

This article is a tutorial which introduces the reader to the theory of rough sets. Rough sets are based on the indiscernibility concept. Acids, for instance are not considered to have significantly different acidity by most chemists if their pK's are 4.03 and 4.14. They will usually classify acids in categories such as very weak, weak, etc. Classification on the basis of such rough descriptions can be performed on the basis of rough set theory. An example from QSAR is given.

   K. De Braekeleer, A. de Juan and D.L. Massart.
"Purity assessment and resolution of tetracycline hydrochloride samples analysed using HPLC with diode array detection". J. Chromatography A, 832, 67-86 (1999).

A tetracycline hydrochloride sample was analyzed using High Performance Liquid Chromatography with Diode Array Detection (HPLC DAD). The HPLC DAD chromatograms were investigated using chemometric methods such as the Orthogonal Projection Approach (OPA), the Fixed Size Moving Window Evolving Factor Analysis Approach (FSW-EFA), Multivariate Curve Resolution Alternating Least Squares Approach (MCR-ALS) applied on the single HPLC DAD chromatograms (two-way data) and on sets of such chromatograms (three-way data).

   C. Vannecke, S. Baré, M. Bloomfield and D.L. Massart.
"An experimental design approach to the optimisation of a flow injection analysis method for glycine". J. Pharm. Biomed. Anal., 18, 963--973 (1999).

A flow injection analysis method for the determination of glycine, based on the reaction with ortho-phthalaldehyde and N-acetylcysteine in a basic buffer, was optimised using experimental design. In the screening step, the use of a half-fraction factorial design was compared with a quarter-fraction factorial one. To determine the significance of the factor effects, different criteria, as (i) a critical effect based on a t-test, (ii) a non parametric method, (iii) a normal probability plot, and (iv) a randomization test method, were compared.

   W. Wu, Q. Guo, D. Jouan-Rimbaud and D.L. Massart.
"Using contrasts as data pretreatment method in pattern recognition of multivariate data". Chem. Intell. Lab. Syst., 45, 39-53 (1999).

A contrast method originally proposed by Spiegelman [C.H. Spiegelman, presented at Compana '95, Wuerzburg, Germany.] is modified to pretreat multivariate data for classification. Three NIR data sets and one pollution data set are used as examples. Our results show that the contrast method greatly improves the ratios of between- to within-class variance. This conclusion does not depend on the type of classifier used. Regularised discriminant analysis and partial least squares with univariate feature selection based on Fisher's ratio were applied here.

   K. De Braekeleer, R. De Maesschalck, P.A. Hailey, D.C.A. Sharp and D.L. Massart.
"On-line application of the orthogonal projection approach (OPA) and the soft independent modelling of class analogy approach (SIMCA) for the detection of the end point of a polymorph conversion reaction by NIR". Chem. Intell. Lab. Syst., 46, 103-116 (1999).

Near infrared spectra were recorded during a polymorph conversion of P1 into P2. The spectra were measured on-line in the reaction vessel, which was held at a constant temperature. The dissimilarity criterion used in the orthogonal projection approach (OPA) was evaluated to follow the reaction progress and to detect the end point of the conversion process on-line. The soft independent modelling of class analogy approach (SIMCA) was also investigated for the detection of the start and the end point of the reaction.

   Q. Guo, W. Wu and D.L. Massart.
"The robust normal variate transform for pattern recognition with near-infrared data". Anal. Chim. Acta, 382, 87-103 (1999).

The standard normal variate transform (SNV) is applied to pretreat NIR data for pattern recognition. The results show that SNV improves classification results in most of the cases by reducing the within-class variance. Because of the closure problem, SNV leads to artefacts and is difficult to interpret in simple methods of univariate direct discrimination. A proposed robust normal variate trancform (RNV) gives more reasonable results than SNV. The drawback of RNV is that the applied percentile needs to be optimised. A proposal for quick selection of the percentile is given.

   R. De Maesschalck, A. Candolfi, D.L. Massart and S. Heuerding.
"Decision criteria for SIMCA applied to near infrared data". Chem. Intell. Lab. Syst., 47, 65-77 (1999).

The class boundaries in SIMCA can be based on different distances, such as the Euclidean, the Gnanadesikan and the Mahalanobis distance. The latter led to more robust classifications for the data set investigated here (NIR spectra of four categories of tablets). To decide whether a new object belongs to the class or not, the original scores are replaced by scores predicted by a leave-one-out procedure. In this way less objects that do belong to the class are (wrongly) rejected as not belonging to it.

Y. Vander Heyden, J. Saevels, E. Roets, J. Hoogmartens, D. Decolin, M.G. Quaglia, W. Van den Bossche, R. Leemans, O. Smeets, F. Van de Vaart, B. Mason, G.C. Taylor, W. Underberg, A. Bult, P. Chiap, J. Crommen, J. De Beer, S.H. Hansen, D.L. Massart.

"Interlaboratory studies on two HPLC assays for tylosin (tartrate)". J. Chromatogr. A, 830, 3-28 (1999).

An interlaboratory study was performed on two HPLC methods for tylosin, a reversed phase HPLC method on a C-18 column and one using a polymeric stationary phase. The first method is described in pharmacopoeia monographs, to determine the composition of a tylosin mixture, while the second is proposed for both the composition and the content determination. The interlaboratory studies were set-up and interpreted according to ISO-guidelines. The paper is written as a tutorial article explaining the principles and methods of these guidelines. The results of both methods were compared. Both were found to have disadvantages but in general the old method still can be preferred, both for composition determination as to assay the components.

   Y. Vander Heyden, F. Questier and D.L. Massart.
"Ruggedness testing of chromatographic methods : selection of factors and levels". J. Pharmac. Biomed. Anal., 18, 43-56 (1998).

The first step in a ruggedness test is the selection of factors and their levels. Both topics are discussed, completing a strategy described earlier. It is demonstrated that depending on the definition of a factor, information, physically more or less meaningful, is extracted from the experimental design results. A guideline to select the levels of the factors in a ruggedness test was proposed. Some special cases, i.e. asymmetric intervals around the nominal level, were also discussed.

J.J.N. Cirunay, Y. Vander Heyden and J. Plaizier-Vercammen.

"Separation from Related Compounds and Assay of Calcipotriol by High-Performance Liquid Chromatography". Journal of Chromatographic Science, 36, 417-421 (1998).

Two methods are presented. One involves the separation of calcipotriol from two related compounds, cholecalciferol (Vitamin D₃) and calcitriol (1,25-dihydroxyvitamin D₃). The other involves the isolation and assay of calcipotriol from a topical ointment. The study was performed with RP-HPLC using an RP18 column and UV detection. A mobile phase mixture containing methanol-acetonitrile-water (67:23:10, v/v) was found which achieved separation within 18 min. A mobile phase of methanol/water (80:20, v/v) attained a slower elution of calcipotriol. For isolation and assay of calcipotriol from an ointment, dissolution in chloroform gave the highest recovery (>98%). The isolation and assay process can be performed within 2 h.

   F. Questier, Y. Vander Heyden and D.L. Massart.
"RTS, a computer program for the experimental set-up and interpretation of ruggedness tests". J. Pharmac. Biomed. Anal., 18, 287-303 (1998).

A program for the experimental set-up and interpretation of ruggedness tests is described. The implemented strategy was based on a number of case studies. The use ot randomization tests as an alternative statistical interpretation method for the significance of effects also was examined. Some of the designs are expandable to designs with given statistical characteristics. The program is designed to facilitate the selection of the designs, the interpretation of the results and to prevent or detect problems such as drifting of responses.

   V. Centner, O.E. de Noord, D.L. Massart.
"Detection of nonlinearity in multivariate calibration". Anal. Chim. Acta, 376, 153-168 (1998).

The present study compares several methods to detect the presence of nonlinearity in multivariate calibration. The explored techniques are applied to four experimental near infrared (NIR) data sets. Mallows augmented partial residual plot is recommended as the most universal diagnostic plot to detect nonlinearity. The significance of nonlinearity is evaluted using the runs test.

   W. Wu, Q. Guo, P.F. de Aguiar, D.L. Massart.
"The star plot : an alternative display method for multivariate data in the analysis of food and drugs". J. Pharmac. Biomed. Anal., 17, 1001-1013 (1998).

The star plot (SP) is a method of displaying multivariate data. Combined with principal component analysis (PCA), more than two PCs can be displayed in one plane. Different varients of this method are applied to an atomic absorption spectrometry (AAS) data set and to three near infra-red (NIR) spectral data sets.

M. Baret, P. Fabry, D.L. Massart, C. Menardo and M. Fraysse.

"Calibration procedures for a halide ion-selective electrode array". Analusis, 26, 267-277 (1998).

Univariate and multivariate chemometric methods are compared for the determination of ion concentrations measured with a 6 Ion-Selective Electrode array. Principal component analysis is applied as an exploratory method to observe the extent to which interferences occur. It is shown that univariate inverse calibration performs at least as well as classical calibration. The use of the former is therefore recommended. The use of partial least squares, principal component regression and multivariate regression is discussed.

   J. Verdú-Andrés, D.L. Massart.
"Comparison of prediction- and correlation-based methods to select the best subset of principal components for principal component regression and detect outlying objects". Appl. Spectroscopy, 52 (11), 1425-1434 (1998).

The use of prediction and correlation criteria for the best subset selection of principal components for principal component regression is compared. Results for both methodologies are similar, and always equal to or better than those obtained by using top-down principal component regression. In this comparison, the prediction criterion is based on the use of leverage-corrected residuals.

   F. Despagne, D.L. Massart.
"Tutorial review : neural networks in multivariate calibration". The Analyst, 123, 157R-178R (1998).

In this tutorial, the basic principles and capabilities of backpropagation neural networks are described, in order to situate them with respect to other chemometrics techniques such as PLS or PCR. The concept of bias and variance familiar to people experienced with soft models is presented in the framework of neural computation. It is demonstrated how these two quantities can be balanced in a neural network. A methodology intended for analytical chemists to model multivariate calibration data with neural networks is then proposed. It is illustrated with some applications and advances reported in quantitative chemical data analysis in the last few years. Some important limitations of neural networks are emphasised. Common issues such as input variable selection or topology optimisation are discussed in detail, and some recent methods to improve interpretability of the models are presented.

   R.J. Poppi and D.L. Massart.
"The optimal brain surgeon for pruning neural network architecture applied to multivariate calibration". Anal. Chim. Acta, 375, 187-195 (1998).

The optimal brain surgeon (OBS) pruning procedure for automatic selection of the optimal neural network architecture was applied in multivariate calibration studies of near infrared data sets. These data sets were preprocessed by using principal component analysis, and the scores of these principal components were the input into the neural network. The pruning procedure improved the generalization ability, reducing the errors in a test set when compared to a non-pruned architecture, and produced better results than PCR and PLS. When using OBS in a network with both linear and non-linear transfer functions, a diagnostic for non-linearity results. In case of a linear model, the net is automatically reduced to principal component regression (PCR).

   V. Centner, D.L. Massart.
"Optimisation in locally weighted regression". Anal. Chem., 70 (19), 4206-4211 (1998).

The application of locally weighted regression (LWR) to nonlinear calibration problems and strongly clustered calibration data often yields more reliable predictions than global linear calibration models. This study compares the performance of LWR that uses PCR and PLS regression, the Euclidean and Mahalanobis distance as a distance measure, and uniform and cubic weighting. Recommendations are given on how to apply LWR to near-infrared data sets.

   A. de Juan, S.C. Rutan, R. Tauler, D.L. Massart.
"Comparison between the direct trilinear decomposition and the multivariate curve resolution alternating least squares methods for the resolution of three-way data sets". Chemom. Intell. Lab. Syst., 40, 19-32 (1998).

Direct trilinear decomposition (DTD) is based on the resolution of the generalized eigenvector/eigenvalue problem, multivariate curve resolution alternating least squares (MCR-ALS) is focused on the optimization of initial estimates by using data structure and chemical constraints. They have been tested on a variety of three-way simulated data sets having common sources of variation in real response profiles, such as signal shift, broadening or shape distortions caused by noise. The effect of these factors has been evaluated. Conclusions inferred from the simulated examples help to clarify the performance of both methods on a real example and to provide some general guidelines to understand better the potential of each method.

   B. Walczak, D.L. Massart.
"Multiple outlier detection revisited". Chem. Intell. Lab. Syst., 41, 1-15 (1998).

   K. Szczubialka, J. Verdú-Andrés, D.L. Massart.
"A new method of detecting clustering in the data". Chem. Intell. Lab. Syst., 41, 145-160 (1998).

   F. Despagne, D.L. Massart.
"Variable selection for neural networks in multivariate calibration". Chem. Intell. Lab. Syst., 40, 145-163 (1998).

The problem of variable selection for neural network modeling is discussed in this paper. Two methods that gave the best results in a previous comparative study are presented. One of these methods is a modified version of the Hinton diagrams, the other method is based on saliency estimation and is part of the Optimal Brain Surgeon algorithm for pruning unimportant weights in a neural network. We also propose two new methods, based on the estimation of the contribution of each input variable to the variance of the predicted response. These new methods are designed for situations where input variables are orthogonal, such as the PC scores often used in multivariate calibration. The four methods are tested on synthetic examples, and on real industrial data sets for multivariate calibration. The main characteristics of each method are discussed. In particular, we underline the strong theoretical and experimental limitations of methods like the modified Hinton diagrams, based on weight magnitude estimation. We also demonstrate that, although the saliency estimation approach is theoretically more stringent, it gives unstable results on repeated trials. The advantage of the two variance-based approaches is that they are much less dependent on the initial weight randomization than the two other methods, and therefore the results they produce are more stable and reliable.

   D. Jouan-Rimbaud, D.L. Massart, C.A. Saby, C. Puel.
"Determination of the representativity between two multidimensional data sets by a comparison of their structure". Chem. Intell. Lab. Syst., 40, 129-144 (1998).

   C. Hartmann, J. Smeyers-Verbeke, D.L. Massart, R.D. McDowall.
"Validation of bioanalytical chromatographic methods". J. Pharm. Biomed. Anal., 17, 193-218 (1998).

This review article discusses a strategy for the validation of chromatographic methods for the quantification of drugs in biological matrices. Both the validation terminology and the hypothesis testing are briefly reviewed. The emphasis is on the design of the experiments required to allow a reliable conclusion about acceptance or rejection of the bioanalytical method.

   L. Pasti, D. Jouan-Rimbaud, D.L. Massart, O. de Noord.
"Application of Fourier transform to multivariate calibration of near-infrared data". Anal. Chim. Acta, 364, 253-263 (1998).

The first fifty power-spectra coefficients are obtained and a genetic algorithm is applied to select the coefficients that yield the best multiple regression model. The method is applied to the multivariate calibration of the hydroxyl number in a polyol.

W. Prus, Y. Vander Heyden, D.L. Massart and T. Kowalska.

"Modelling of solute retention in normal-phase HPLC with the chemically bonded 3-cyanopropyl stationary phase". Acta Chromatographica, 8, 98-107 (1998).

   Frédéric Despagne, Beata Walczak, D.Luc Massart.
"Transfer of calibrations of near-infrared spectra using neural networks". Appl. Spectroscopy, 52 (5), 732-745 (1998).

An approach based on the use of neural networks is proposed. The proposed method builds a single transfer model for all spectral windows. The fixed size spectral windows and their position indices constitute the training objects. No prior background correction is required. On the studied data sets, the neural network approach was found to perform at least as well as PDS for both reconstruction and calibration.

   R. De Maesschalck, F. Cuesta-Sánchez, D.L. Massart, P. Doherty, P. Hailey.
"On line monitoring of powder blending with near infrared spectroscopy". Appl. Spectroscopy, 52 (5), 725-731 (1998).

Powder blending was monitored on-line by taking near-infrared measurements during the process. The average standard deviation between the measurements taken at each time and the dissimilarities with an ideal spectrum are used for monitoring purposes. SPC charts, both univariate and multivariate, were developed for monitoring characteristic wavelengths.

   E. Bouveresse, C. Casolino, D.L. Massart.
"Assessing the validity of near-infrared monochromator calibrations over time". Appl. Spectroscopy, 52 (4), 604-612 (1998).

Validation samples are analysed at regular time intervals by NIR and the reference method. When unacceptable differences are detected, it is necessary to identify the source of the errors and a strategy is proposed to achieve this. In simple cases the strategy also allows correction of the error.

   K. De Braekeleer, F. Cuesta-Sànchez, P.A. Hailey, D.C.A. Sharp, A.J. Pettman, D.L. Massart.
"Influence and correction of temperature perturbations on NIR spectra during the monitoring of a polymorph conversion process prior to self-modelling mixture analysis". J. Pharm. Biomed. Anal., 17, 141-152 (1998).

A polymorph conversion of species P1 to P2 was monitored on-line by near infrared spectroscopy. Two sets of data measured at constant and varying temperature were considered for the same process. In order to obtain a better comparison of the resolution results a temperature correction was proposed. The resolution of the spectral data in the concentration profiles and the spectra was obtained with SIMPLISMA and OPA.

   Y. Vander Heyden, F. Questier, D.L. Massart.
"A ruggedness test strategy for procedure related factors : experimental set-up and interpretation". J. Pharm. Biomed. Anal., 17, 153-168 (1998).

A strategy to perform ruggedness tests is described. The different steps in the set-up and in the interpretation of the results are given. The strategy is based on a number of case studies, allow a statistical interpretation of the effects and was implemented in a software tool. The strategy was completed with a number of minimal screening designs which reduce the number of experiments required but in consequence only allow a limited or no statistical interpretation of the effects. Some of the minimal designs are expandable to designs with characteristics similar to those in the initial strategy.

H. Spapen, Y. Vander Heyden, M. Diltoer, J. Ramet, D.L. Massart, L. Huyghens.

"N-acetyl-L-cysteine (NAC) in early human septic shock : plasma levels and effects on proinflammatory cytokines". Am. J. Respiratory and critical care medicine, 157 (3), A296 (1998).

   V. Centner, D.L. Massart, S. de Jong.
"Inverse calibration predicts better than classical calibration". Fres. J. Anal. Chem., 361, 2-9 (1998).

Inverse calibration, i.e. concentration = f(measurement) yields more reliable predictions than classical calibration, i.e. measurement = f(concentration). The improvement increases with decreasing precision of the measurements.

   A. Candolfi, W. Wu, D.L. Massart, S. Heuerding.
"Comparison of classification approaches applied to NIR-spectra of clinical study lots". J. Pharm. Biomed. Anal., 16, 1329-1347 (1998).

Pattern recognition is applied to NIR spectra of clinical study lots. Linear discriminant analysis, quadratic discriminant analysis and K-nearest neighbour methods are compared. It is investigated if several batches can be merged in one class or whether it is necessary to model each batch separately.

   S. Kuttatharmmakul, D.L. Massart, J. Smeyers-Verbeke.
"Influence of precision, sample size and design on the beta error of linearity tests". J. Anal. Atomic Spectrometry, 13, 109-118 (1998).

A straight-line calibration model is often applied in the analysis. Before its use in routine, the suitability of the straight-line model in the calibration procedure has to be evaluated. Therefore a linearity test must be performed to check whether the data are well fitted by a straight line. In this article the suitability of linearity tests in the validation of a straight line calibration model from a restricted number of measurements (n equals 6 to 10) is studied. Two frequently used linearity tests are evaluated in this work namely, the anova lack-of-fit to the first degree model and the test of significance of b2, the second degree coefficient from the second degree model fitted through the data. For the detection of lack-of-fit to the straight-line model, the test of significance of b2 performs better than the ANOVA lack-of-fit test. The design with 3 concentration levels gives the best result. The lower the measurement precision, the higher the beta error. Heteroscedasticity largely increases the probability that a lack-of-fit is not detected. The formulae to determine i) the sample size that allows the detection of a certain deviation (from linearity) with a specified beta error and ii) the beta error for a given experimental design were derived and evaluated. The results obtained from the formulae correspond with those obtained from the data simulation.

M.I. Griep, E. Bory, K. Collys, D.L. Massart.

"Category ratio scale as an alternative to magnitude matching for age-related taste and odour perception". Food Quality and Preference, 9 (1/2), 67-72 (1998).

   K. De Braekeleer, D.L. Massart.
"Evaluation of the orthogonal projection approach (OPA) and the Simplisma approach on the Windig standard spectral data sets". Chemom. Intell. Lab. Systems, 39, 127-141 (1997).

The advantages and limitations of the orthogonal projection approach (OPA), in comparison with the Simple-to-use interactive self-modeling mixture analysis approach (SIMPLISMA), are investigated on the standard spectral data published by Windig. These data (four different industrial sets) have been proposed in Chemometrics and Intelligent Laboratory Systems as standard data for the evaluation of self-modeling methods that extract information from bilinear data matrices.

M. Griep, T.F. Mets, K. Collys, D. Verté, G. Verleye, I. Ponjaert, D.L. Massart.

"MNA and odour perception". In : MNA in the elderly; Raven Press Lippincott, New York, pp 1-22 (1997).

Many studies show that odour perception declines with age while diminisished odour perception is associated with poor general health, malnutrition and various diseases. Food flavours consist mainly of volatile odours and flavour perception has been rated as the strongest determinant of food choice in the elderly. Enriching food by flavours to compensate for diminished odour perception may therefore improve consumption of particular food and prevent malnutrition.
In this study, it is explored which age-associated factores such as poor oral health, body composition, diminished odour perception and poor general health are accompanied by increased risk for malnutrition, measured by the MNA.

   D. Jouan-Rimbaud, B. Walczak, R.J. Poppi, O.E. de Noord, D.L. Massart.
"Application of wavelet transform to extract the relevant component from spectral data for multivariate calibration". Anal. Chem., 69, 4317-4323 (1997).

An approach aiming at extracting the relevant component for multivariate calibration is introduced for the modeling of near-infrared data. The extraction of the relevant component is carried out in the wavelet domain. The PLS results on these relevant features are better, and therefore, it seems that this approach can successfully be used to remove noise and irrelevant information from speclra for multivariate calibration.

P.F. de Aguiar and D.L. Massart.

"Experimental design", Encyclopedia of Computational Chemistry, John Wiley and Sons, 1997.

Tutorial

P.F. de Aguiar, M. Jimidar, D.L. Massart.

"Application of the logistic transformation for prediction of anion mobility in capillary ion analysis". Acta Chromatographica, 7, 35-48 (1997).

The logistic transformation has been applied in capillary zone electrophoresis to model mobility as a function of pH. The results have heen compared with those obtained by a non-linear regression procedure.

   P.F. de Aguiar, B. Bourguignon, D.L. Massart.
"Comparison of models and designs for optimisation of the pH and solvent strength in HPLC". Anal. Chim. Acta, 356, 7-18 (1997).

The performance of a design selected with the algorithm of Kennard and Stone and a D-optimal design were compared. As a practical example the separation of a mixture of β-blockers is reported.

   B. Walczak, D.L. Massart.
"Wavelets - something for analytical chemistry ?". TrAC, 16 (n°8), 451-463 (1997).

Wavelets have shown great applicability in many diverse fields of science, and are now becoming of interest in analytical chemistry. This paper is intended as a first reading, introducing fundamentals of wavelets and wavelet transforms, and some applications thereof, such as signal compression and denoising, image processing, data set compression and the modeling of multivariate data sets.

   F. Despagne, D.L. Massart, O.E. De Noord.
"Optimization of partial-least-squares calibration models by simulation of instrumental perturbations". Anal. Chem., 69, 3391-3399 (1997).

A critical step in partial-least-squares (PLS) modeling is the model optimization. Cross-validation is often applied, but in spite of its statistical properties, it suffers some severe shortcomings. In particular, cross-validation has a tendency to give overfitted models, whereas parsimonious models should be preferred. We propose an alternative form of internal validation, based on the simulation of instrumental perturbations on a subset of calibration samples. A simple criterion is proposed for the adjustment of perturbations. The method is applied for the validation of nine PLS1 calibration models on industrial data sets and compared with cross-validation and cross-validation combined with a randomization test. It is shown that parsimonious models can be obtained, with a good predictive power when they are applied to external test data.

M.I. Griep, T.F. Mets, K. Collys, P. Vogelaere, M. Laska and D.L. Massart.

"Odour perception in relation to age, general health, anthropometry and dental state". Arch. Gerontology and Geriatrics, 25, 263-275 (1997).

Considering the possible role of age-related phenomena such as general health, dental health and nutrition in odour perception, their joint effect on variability in odour perception was evaluated in the present study. The SENIEUR protocol was used to assess the general health status to assess the nutritional status of the persons. The sensory detection threshold for isoamyl acetate was determined to measure odour perception. Oral examinations were used to assess dental status. Those in poor general health had lower mean values for odour perception than those in good or reasonably general health. Complete denture wearers and those with partial dentures had significantly lower values for odour perception than those with only natural teeth. Odour perception correlated significantly with measures of body fat and muscle mass. Our results indicate that general health and dental state are important age-associated factors in odour perception.

M.I. Griep, P. Van der Niepen, J.J. Sennesael, T.F. Mets, D.L. Massart and D.L. Verbeelen.

"Odour perception in chronic renal disease". Nephrology Dialysis Transplantation, 12, (n°10), 2093-2098 (1997).

Our results show that the ability to smell is severely impaired in patients with chronic renal failure and is related to the degree of renal impairment and the degree of accumulation of uraemic toxins. After renal transplantation, patients have a normal odour perception, indicating the capacity of the olfactory system to recover once the concentration of uraemic toxins remains below a critical threshold. Acute removal of uraemic toxins by dialysis does not correct olfactory disturbances, suggesting a long lasting effect of uraemia on olfactory function.

W. Prus, Y. Vander Heyden, P. Kus, D.L. Massart, T. Kowalska.

"A statistical evaluation of selected RPHPLC retention models". Acta Chromatographica, 7, 194-209 (1997).

Three RP-HPLC models for solute retention proposed by Snyder, Schoenmakers and Kowalska respectively, were statistically evaluated using six crown ethers as test solutes and binary methanol-water mixtures as mobile phases. Five RP columns (C8 and C18) were tested. The model proposed by Schoenmakers was found to provide the best predictions for the retention parameter, lnk'.

   P.F. de Aguiar, Y. Vander Heyden, D.L. Massart, R. Leardi, J.O. De Beer.
"Optimisation of the isocratic separation of dithranol and related compounds by reversed-phase liquid chromatography". Acta Chromatographica, 7, 129-148 (1997).

An optimal mobile phase composition is determined for the resolution of dithranol and four related compounds in an isocratic reversed-phase HPLC run. A combination of Snyder's solvent triangle concept and mixture design was used. The quantification of the five substances was tested and the ruggedness of the separation for small changes in the mobile phase composition was verified.

   B. Walczak, E. Bouveresse, D.L. Massart.
"Standardization of near-infrared spectra in the wavelet domain". Chemom. and Intell. Lab. Systems, 36, 41-51 (1997).

For each NIR spectrum, wavelet transforms are computed, and the transfer step is performed by relating the wavelet transform coefficients of spectra on both instruments with univariate linear models. The results are better than those obtained by PDS. The approach offers advantages only when the spectral differences are local. When they are global, other methods such as the slope/bias method are to be preferred.

   A. Baldovin, W. Wen, D.L. Massart and A. Turello.
"Regularised discriminant analysis (RDA)-modelling for the binary discrimination between pollution types". Chem. Intell. Lab. Systems, 38, 25-37 (1997).

Discrimination between different types of pollution is carried out with RDA. Attention is paid to the four extreme discrimination models that are special cases of RDA and which are the well known linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA), the nearest-means classifier (NM) and the weighted nearest means-classifier (WNM). Feature selection to determine the most important variables for the discrimination is studied.

   B. Walczak and D.L. Massart.
"Wavelet packet transform applied to a set of signals: a new approach to the best-basis selection". Chem. Intell. Lab. Systems, 38, 39-50 (1997).

A new approach to best-basis selection for a set of signals, enabling significant and uniform data compression in the time-frequency domain, is proposed. The high degree of data compression can be used as the first step of fast approximate principal component analysis (PCA) of data sets with high rank.

Y. Vander Heyden, A. Bourgeois, D.L. Massart.

"Influence of the sequence of experiments in a ruggedness test when drift occurs". Anal. Chim. Acta, 347, 369-384 (1997).

When a drift or time effect occurs, the sequence of execution of design experiments can affect the estimated factor and interaction effects. Ageing of a chromatographic column can cause drift in some responses. Different sequences of similar designs were performed. Several factor and interaction effects were found to be confounded with the time effect. The extent of confounding of an estimated effect depends on the sequence in which design experiments were performed. Correction of the results based on replicated nominal experiments yields effects free of the time effect, regardless of the sequence of the performed design experiments.

D.L. Massart.

"Medische en farmaceutische toepassingen van de hoofdcomponentenanalyse". Uit : Verhandelingen van de Koninklijke Academie voor Geneeskunde van België, LIX, n°4, 287-325 (1997).

Medical and pharmaceutical research leads to a collection of data that are presented in the form of tables. Examples from several fields in the medical and pharmaceutical sciences show that principal component analysis allows to visualise and thereby to more easily interpret this information. The method is explained algebraically and geometrically.

   W. Wu and D.L. Massart.
"Regularised nearest neighbour classification method for pattern recognition of near infrared spectra". Anal. Chim. Acta, 349, 253-261 (1997).

When a data set contains a high number of variables compared to the number of objects, the k nearest neighhour classification method (kNN) cannot be applied with the Mahalanobis distance as similarity criterion. To solve this problem, kNN is modified to the regularised nearest neighbour classification method (RNN) by using the regularised covariance matrix in the Mahalanobis distance in the same way that LDA and/or QDA are modified to regularised discriminant analysis (RDA). Our results demonstrate that RNN improves the classification results. RNN performs better than RDA when the normal distribution assumption is severely violated, and worse when the data are normally distributed, since it is a non-parametric version of RDA.

   E. Bouveresse, S.C. Rutan, Y. Vander Heyden, W. Penninckx and D.L. Massart.
"Detection, interpretation and correction of changes in the instrumental response of near-infrared monochromator instruments over time". Anal. Chim. Acta., 348, 283-301 (1997).

A new approach is proposed for the detection of changes in the instrumental response of near-infrared spectrometers over time. This approach is based on simulating possible instrumental differences and performing a principal component analysis of the simulated data set. This multivariate approach is tested on a real near-infrared data set and the results are compared to those obtained by the several univariate diagnostic procedures commonly used to check the stability of the responsc of a near-infrared spectrometer over time. A simulated data set is used to study the influence of using different types of standards on the results obtained by this approach.

   D. Jouan-Rimbaud, D.L. Massart, C.A. Saby, C. Puel.
"Characterization of the representativity of selected sets of samples in multivariate calibration and pattern recognition". Anal. Chim. Acta, 350, 149-161 (1997).

Two statistical tests are proposed, to compare two data sets, and estimate their representativity. The first one is the comparison of the variance-covariance matrices of the two data sets: their equality implies that both data sets have the same direction in space, and that the spread around the mean is similar. Then, the Mahalanobos distance between the centroids of the two sets is calculated, in order to know wether the centroids have the same position.

   P.F. de Aguiar, Y. Vander Heyden and D.L. Massart.
"Study of different criteria for the selection of a rugged optimum in HPLC optimisation". Anal. Chim. Acta., 348, 223-235 (1997).

Four criteria to select acceptably good and robust experimental conditions during optimisation in HPLC are proposed and evaluated. All criteria take into account the separation between the worst separated peak pair and the robustness of the optimisation criterion around each possible optimum. The optimal conditions selected by the criteria were found to be Pareto-optimal and in agreement with those selected applying Derringer's desirability functions.

   J. Verdú-Andrés, D.L. Massart, C. Menardo and C. Sterna.
"Correction of non-linearities in spectroscopic multivariate calibration by using transformed original variables and PLS regression". Analyt. Chim. Acta, 349, 271-282 (1997).

The use of transformed original variables is proposed to correct non-linearities in the original data. The additional information that these new variables introduce can be linearly modelled by the PLS algorithm, giving better models than the ones obtained by using only the original variables. Non-informative absorbances can be eliminated later giving more stable models.

   A. de Juan, Y. Vander Heyden, R. Tauler, and D.L. Massart.
"Assessment of new constraints applied to the alternating least squares (ALS) method". Anal. Chim. Acta, 346, 307-318 (1997).

The introduction of constraints in multivariate curve resolution methods, as the Alternating Least Squares (ALS), is commonly used to limit the span of possible solutions, guiding the iterative process to a final result as close as possible to the true situation. Two modifications of the unimodality constraint and a new constraint for chromatographic concentration profiles related to the prevention of fronting were examined. Simulated and real data sets were used to evaluate the effect of the constraints on the resolution results. The goodness of the constraints was evaluated relative to the recovery of the concentration profiles and to the quality of the data fit.

M. Jimidar, P.F. de Aguiar, S. Pintelon and D.L. Massart.

"Selectivity optimization for the separation of chlorophenols in an irregularly shaped experimental region in capillary electrophoresis". J. Pharm. Biom. Anal., 15, 709-728 (1997).

The separation of seventeen chlorophenol congeners and phenol was studied. The pH and the concentration of sodium dodecylsulphate were found to be important. During the implementation of a central composite design for the optimization of the separation it appeared that a part of the domain was not feasible as it resulted in very long migration times and extremely deformed peaks. Therefore, a D-optimal design was selected within the boundaries of the feasible region. It was possible to find a region for the simultaneous separation of 15 compounds. Further optimization at these optimal conditions resulted in a separation where 17 peaks could be observed.

   W. Wu, D.L. Massart and S. de Jong.
"Kernel PCA algorithms for wide data. Part II. Fast cross-validation and application in classification of NIR data". Chemom. Intell. Lab. Systems., 37, 271-280 (1997).

   W. Wu, D.L. Massart and S. de Jong.
"The kernel PCA algorithms for wide data. Part I : theory and algorithms". Chemom. Intell. Lab. Systems, 36, 165-172 (1997).

Four classic PCA algorithms: NIPALS, the power method (POWER), singular value decomposition (SVD) and eigenvalue decomposition (EVD) are modified into their kernel version to analyse wide data sets. For such data sets with many variables and fewer objects, the classic algorithms become very inefficient, because the size of the associated matrix X' . X (p x p) is very large. Based on the kernel matrix X . X' (n x n), the kernel algorithms are developed. They yield the same principal components but, for wide data sets, they are faster than the corresponding classic algorithms. Simulation results show that the kernel EVD is the most efficient algorithm, and that the difference between the kernel SVD and EVD is not very large.

K.D. Collys, A.C. Roma de Sousa and J. Smeyers-Verbeke.

"Soluble denture adhesives: pH and sodium content." Eur. J. Prosthodont. Rest. Dent., 5 (2), 63-67 (1997).

Seventeen soluble denture adhesives available in local drug stores in Belgium were selected. The pH of aqueous adhesive solutions was determined with a glass pH electrode. Sodium determinations of normal and acidified aqueous adhesive solutions were performed by means of atomic emission spectrometry. The amount of adhesive required per application according to manufacturer's instructions was determined by weight. From this study it was calculated that denture adhesive applications three times a day in both upper and lourer complete denture could result in a supplementary daily sodium intake of 0.9 to 6.0 mmol Na⁺/day.

   C. Hartmann, P. Vankeerberghen, J. Smeyers-Verbeke and D.L. Massart.
"Robust orthogonal regression for the outlier detection when comparing two series of measurement results". Anal. Chim. Acta, 344, 17-28 (1997).

The procedure described is based on the least median of the squared orthogonal residuals (orthogonal LMS). By means of real data sets it is shown that the orthogonal LMS procedure is a useful screening tool to detect outlying values.

   B. Walczak and D.L. Massart.
"Noise suppression and signal compression using wavelet packet transform". Chemom. Intell. Lab. Systems, 36, 81-94 (1997).

The basic concepts of the discrete wavelet transform (DWT) and the wavelet packet transform (WPT) are presented in this tutorial. Different approaches to the WPT coefficients selection aiming at signal compressiopn and denoising are described. Uniform compression of the set of signals is discussed as well.

   S. Ingelbrecht, J.P. Remon, P. Fernandes de Aguiar, B. Walczak, D.L. Massart, F. Van De Velde, P. De Baets, H. Vermeersch and P. De Backer.
"Instrumentation of a roll compactor and the evaluation of the parameter settings by neural networks". Int. J. Pharmaceutics, 148, 103-115 (1997).

A Fitzpatrick L83 Chilsonator was instrumented in order to optimize the roll compaction process using drum-dried waxy maize starch, as a model compound. The interrelation of the four adjustable roll compactor parameter settings namely the velocity of the rolls (RS), the speed of the horizontal (HS) and of the vertical screw (VS), and the air pressure (P_air) influenced the granule quality. As a second order polynomial was not successful to model the behaviour of the friability in function of the four roll compactor parameters, a Multilayer Feed-Forward neural network (MLF) was used. It was shown that the MLF network models the friability more accurately than a second order polynomial.

   F. Cuesta Sanchez, S.C. Rutan, M.D. Gil García and D.L. Massart.
"Resolution of multicomponent overlapped peaks by the orthogonal projection approach, evolving factor analysis and window factor analysis". Chemom. Intell. Lab. Syst., 36, 153-164 (1997).

The orthogonal projection approach (OPA), has been extended by including a step that allows the chromatographic and spectroscopic pure compound profiles to be determined. This is done using an alternating least-squares procedure. The initial estimations are the spectra selected in the first step of OPA. The performance of the OPA algorithm is compared with that of evolving factor analysis (EFA) and window factor analysis (EWFA).

   P.F. de Aguiar, Y. Vander Heyden, Y. Van Oost, T.J. Coomber and D.L. Massart.
"Optimisation of the reversed phase liquid chromatographic separation of atovaquone, proguanil, and related substances". J. Pharm. Biomed. Anal., 15, 1781-1787 (1997).

The separation of the antimalarial drugs, proguanil and atovaquone and six related compounds was obtained in two optimisation steps: a) of the mobile phase composition, and b) of the pH. A mixture design and window selection diagrams were applied. The final conditions allow the detection of the six related compounds down to 0.1%.

   A. Candolfi, D.L. Massart and S. Heuerding.
"Investigation of sources of variance which contribute to NIR-spectroscopic measurement of pharmaceutical formulations". Anal. Chim. Acta, 345, 185-196 (1997).

The variance contributions of the following factors to the NIR-measurement of tablets, hard and soft gelatine capsules are estimated: measurement repeatability, positioning, time, different samples of one batch and different batches. The data analysis is performed on the one hand with PCA (Principal Component Analysis) as a display method and on the other hand numerically with a nested ANOVA design.

   C. Hartmann, J. Smeyers-Verbeke, W. Penninckx and D.L. Massart.
"Detection of bias in method comparison by regression analysis". Anal. Chim. Acta, 338, 19-40 (1997).

This work investigates the risk of not detecting a systematic difference between two analytical methods (=β-error) and determines the number of points required to detect a constant or a proportional bias with a specified β-error. Different designs to distribute the data are studied. A procedure is also proposed to simultaneously test slope and intercept of the orthogonal least-squares line.

M.I. Griep, T.F. Mets, P. Vogelaere, K. Collys, M. Laska and D.L. Massart.

"Geurperceptie in relatie tot leeftijd, algemene gezondheid,voedingstoestand en gebitsstatus". Tijdschr. voor Gerontologie en Geriatrie, 28, 11-17 (1997).

M.I. Griep, T.F. Mets and D.L. Massart.

"Different effects of flavour amplification of foods on preference and consumption in young and elderly subjects". Food Quality and Preference, 8 (2), 151-156 (1997).

Adjusting flavour to compensate for diminished odour perception may improve the quality of life for the increasingly older population. In this study, we test whether flavour amplification can induce changes in the consumption of particular food. The majority of the younger subjects prefer the low flavour soup, Quorn^® and yoghurt, while the elderly subjects prefer the high flavour level soup, Quorn^® and yoghurt. Comparing the consumed quantities of high and low flavour meals shows that for the younger subjects, flavour amplification decreases consumption levels of soup and yoghurt. For the elderly subjects, amplification of yoghurt increases consumed quantities.

   B. Walczak, E. Bouveresse and D.L. Massart.
"Standardization of near-infrared spectra in the wavelet domain". Chemom. Intell. Lab. Syst., 36, 41-51 (1997).

A new standardization method based on transferring NIR spectra in the wavelet domain is proposed. For each NIR spectrum, wavelet transform coefficients are computed, and the transfer step is performed by relating the wavelet transform coefficients of spectra obtained on both instruments with univariate linear models. This method is applied to two different data sets, and its performance is compared with the piecewise direct standardization method. The results obtained with this new method are better than those obtained by PDS, the biggest improvement occurring when very few standardization samples were used to estimate the standardization parameters. Both PDS and wavelet approaches offer advantages only when the spectral differences are local. When spectral differences are global, other methods such as the slope/bias correction function as well.

   E. Bouveresse, C. Sterna, J.L. Linossier and D.L. Massart.
"A simple approach for the standardisation of near-infrared filter instruments from monochromator instruments". Analusis, 24, 394-397 (1996).

In process analytical chemistry, standardisation of near-IR spectrometric instruments has become a necessary step, in order to avoid time-consuming calibration procedures. However, most of the existing standardisation methods can only deal with instruments of similar resolution. A simple approach is proposed in order to standardise filter instruments from monochromator instruments, and this approach is tested on a NIR data set of cellulose acetate samples.

   W. Wu, S.C. Rutan, A. Baldovin and D.L. Massart.
"Feature selection using the Kalman filter for classification of multivariate data". Anal. Chim. Acta, 335, 11-22 (1996).

A Kalman filter is developed as a feature selection method and classifier for multivariate data. Three near-infrared (NIR) data sets and a pollution data set are analyzed. For the two most difficult data sets, the Kalman filter successfully selects the wavelengths which lead to very good results with a correct classification rate (CCR) equal to one. These results are much better than the best results obtained from regularized discriminant analysis (RDA) using Fourier transform (FT), principal component regression (PCA). The disadvantage of the Kalman filter is that it needs more memory and more computing time.

   Y. Vander Heyden, C. Hartmann, A.J.M. Hollands, D.L. Massart, P.Nuyten and P. Schoenmakers.
"Ruggedness testing of a size exclusion chromatographic assay for low-molecular-mass polymers". J. Chromatography A., 756, 89-106 (1996).

A ruggedness test of a size-exclusion chromatographic assay for a low-molecular-mass polymer mixture was performed. Eight factors were examined in a reflected fractional factorial design. The factors column manufacturer and detector type turned out to be the least robust. Variations of the other factors, in the examined intervals, do not have chromatographically relevant consequences.

   S.C. Rutan, E. Bouveresse, K.N. Andrew, P.J. Worsfold and D.L. Massart.
"Correction for drift in multivariate systems using the Kalman filter". Chemom. Intell. Lab. Systems, 35, 199-211 (1996).

A approach for detection and correction of time-dependent drift in multivariate calibration parameters is described. Kalman filters can be devised that correct for drift in the baseline and drift in sensitivities, either through purely random processes, or a combination of linear and random drift. Some examples of implementations of this filter are demonstrated using synthetic, near-infrared, and UV - visible spectrophotometric data. The application of the Kalman filter to the inverse calibration problem is also demonstrated.

   D. Jouan-Rimbaud, D.L. Massart and O.E. de Noord.
"Random correlation in variable selection for multivariate calibration with a genetic algorithm". Chemom. Intell. Lab. Systems, 35, 213-220 (1996).

The importance of the validation step in multiple linear regression of near-infrared spectroscopic data, after selection of wavelengths by genetic algorithm, is investigated with the use of random variables. In spite of a careful validation procedure, the GA can select irrelevant variables. The effect is reduced by applying a forward selection in the subsets selected.

   Y. Vander Heyden and D.L. Massart;
"Review of the use of robustness and ruggedness in Analytical Chemistry. Chapter 3". In : Robustness of analytical chemical methods and pharmaceutical technological products; Vol. 19. Data handling in science and technology; Eds. M.M.W.B. Hendriks, J. De Boer; A.K. Smilde; Elsevier, 1996.

The review describes the use of robustness and ruggedness in analytical chemistry. The different definitions for ruggedness and robustness were considered. The consecutive steps in robustness testing are treated and the different alternatives for each step were discussed. Finally a number of case studies taken from the literature were described.

M.I. Griep, G. Verleye, A.H. Franck, K. Collys, T.F. Mets and D.L. Massart.

"Variation in nutrient intake with dental status, age and odour perception". Europ. J. Clin. Nutrition, 50, 816-825 (1996).

Since the proportion of elderly with an insufficient intake of nutrients is high and many of the elderly have poor odour perception or poor dental state, in this study, the relation between age, odour perception, dental state and nutrient intake is explored. Although odour perception and dental state can not fully explain variability in nutrient intake, our results show that people with poor odour perception have lower nutrient intake levels than people with good odour perception. Dental state may not be a direct cause of poor nutrition but a contributing factor in those elderly who have other risk factors.

   W. Wu and D.L. Massart.
"Artificial neural networks in classification of NIR spectal data : selection of the input". Chemom. Int. Lab. Systems., 35, 127-135 (1996).

Artificial neural networks (NN) with back-error propagation are used as a classifier to identify different strengths of drugs in formulations, different qualities of solvents and polymers through NIR spectra. Several pretreatment methods - univariate feature selection, principal component analysis (PCA) and Fisher transformation (FIT) and some combinations - are developed. PCA followed by FIT greatly reduces the architecture of NN for eight of the nine investigated data sets. Our results suggest that PCA/FIT is a useful way to pretreat the data as an input for NN.

   P. Vankeerberghen, J. Smeyers-Verbeke and D.L. Massart.
"Description and implementation of the system for the run suitability check and explorative method validation". Lab. Automat. Inform. Management, 32, 87-102 (1996).

The goal of this paper is to describe
i) the external representation of the decision support system for the run suitability check and explorative method validation in atomic absorption spectrometry by means of representative examples.
ii) the implementation of the whole system. It is explained how the authors came to the hybrid architecture of the system.

   F. Cuesta Sánchez, B. van den Bogaert, S.C. Rutan and D.L. Massart.
"Tutorial : multivariate peak purity approaches". Chem. Intell. Lab. Systems, 34, 139-171 (1996).

The techniques described are applicable to bilinear data matrices obtained from evolutionary processes, i.e., the concentration of each compound evolves with an ordered variable such as time or pH. Their performance for the detection of a minor compound in the presence of a main one is discussed. The application to systems with more than 2 compounds and the effect of smoothing are also considered.

   R. Gargallo, F. Cuesta Sánchez, A. Izquierdo-Ridorsa and D.L. Massart.
"Application of eigenstructure tracking analysis and simplisma to the study of the protonation equilibra of cCMP and several polynucleotides". Anal. Chem., 68 (13), 2241-2247 (1996).

The application of eigenstructure tracking analysis (ETA) and SIMPLISMA for the investigation of the protonation equilibria is proposed. Both approaches have been applied in the pH and in the wavelength direction to the spectroscopic data matrices obtained in the study of each equilibrum. ETA provides information about the number of components in the system, their evolution along the titration, and the local rank. SIMPLISMA is used to obtain the number of compounds in the system, the concentration profiles, and the unit spectrum of each compound.

M. Jimidar, B. Bourguignon and D.L. Massart.

"Application of Derringer's desirability function for the selection of optimum separation conditions in capillary zone electrophoresis". J. Chromatography A, 740, 109-117 (1996).

In order to obtain a set of optimal experimental conditions for the separation of rare earth metal ions in capillary electrophoresis (CE), a multicriteria approach is applied. For this purpose, the Derringer's desirability function is proposed to determine conditions that will result in the most desirable combination of separation, sensitivity and analysis time.

   W. Wu, Y. Mallet, B. Walczak, W. Penninckx, D.L. Massart, S. Heuerding and F. Erni.
"Comparison of regularized discriminant analysis, linear discriminant analysis and quadratic discriminant analysis applied to NIR data". Anal. Chim. Acta, 329, 257-265 (1996).

Three classifiers, namely linear discriminant analysis (LDA), quadratic discriminant analysis (QDA) and regularized discriminant analysis (RDA) are considered in this study for classification based on NIR data. A feature selection method is used to reduce the data dimensionality, and the selected features were used as the input of the classifiers. RDA can be considered as an intermediate method between LDA and QDA, and in several cases, RDA reduces to either LDA or QDA depending on which is better. It is concluded that in many cases, LDA or QDA should be recommended for practical use. However, in those cases where small gains in classification quality are important, the application of RDA might be useful.

   V. Centner, D.L. Massart and O. De Noord.
"Detection of inhomogeneities in sets of NIR spectra". Anal. Chim. Acta, 330, 1-17 (1996).

Methods for the detection of outliers and clusters in data sets for multivariate calibration are discussed. The Cochran test applied to the sum of absorbances is shown to identify outlying replicates. Sample outliers were detected on latent variables (PCs) or by using Rao's statistics combined with Grubbs' and Dixon's tests. The tendency to cluster was eavluated by Hopkins statistic.

   B. Walczak and D.L. Massart.
"The radial basis functions - Partial least squares approach as a flexible non-linear regression technique". Anal. Chim. Acta, 331, 177-185 (1996).

A new approach founded on Radial Basis Functions and Partial Least Squares is proposed to model non-linear chemical systems. Its performance is demonstrated and compared with those of Multilayer Feedforward Network, Radial Basis Function Network, and Spline-PLS. Good performance and a guaranteed learning algorithm of RBF-PLS approach makes it an attractive alternative for earlier established methods.

   B. Walczak and D.L. Massart.
"Application of radial basis functions - Partial least squares to non-linear pattern recognition problems: diagnosis of process faults". Anal. Chim. Acta, 331, 187-193 (1996).

Performance and robustness of a newly proposed approach (based on the Radial Basis Function and PLS2) in the non-linear pattern recognition problem is studied and compared with those of Radial Basis Function Network and multilayer feed-forward network. An example concerns classification of process faults.

   A. Baldovin, W. Wu, V. Centner, D. Jouan-Rimbaud, D.L. Massart, L. Favretto, A. Turello.
"Feature selection for the discrimination between pollution types with partial least squares modelling". Analyst, 121, 1603-1608 (1996).

Discrimination between different types of pollution was carried out with partial least squares (PLS). The data sets concern steelworks and sludges. A method of feature selection is proposed and satisfactory classification rates were achieved. The results show that feature selection can improve the performance of the PLS classification and reduce the number of analytical determinations to be performed.

   R. Gargallo, R. Tauler, F. Cuesta-Sanchez and D.L. Massart.
"Validation of alternating least squares multivariate curve resolution for chromatographic resolution and quantitation". TrAC, 15 (7), 279-286 (1996).

The application of multivariate curve resolution using alternating least-squares (ALS) method for the resolution and quantitation of mixtures of two co-eluted compounds is validated for hydrocortisone (main compound) and prednisone (impurity) in liquid chromatography with diode array detection. Initial estimates are provided by evolving factor analysis and SIMPLISMA methods.

   V. Centner, D.L. Massart, O.E. de Noord, S. de Jong, B.M. Vandeginste and C. Sterna.
"Elimination of uninformative variables for multivariate calibration". Anal. Chem., 68 (21), 3851-3858 (1996).

A method for the elimination of uninformative variables in multivariate data sets is proposed. Artificial (noise) variables are added and a closed form of the PLS or PCR model is obtained for the data set containing the experimental and the artificial variables. The experimental variables that do not have more importance than the artificial variables, as judged from a criterion based on the b coefficients, are eliminated. Practical aspects are discussed on experimentally obtained near-IR data sets. It is concluded that the elimination of uninformative variables can imoprove predictive ability.

   W. Wu, B. Walczak, W. Penninckx and D.L. Massart.
"Feature reduction by Fourier transform in pattern recognition of NIR data". Anal. Chim. Acta, 331, 75-83 (1996).

Fourier transform coefficients of NIR spectra are used as input to regularised discriminant analysis. Good results are reported

M.I. Griep, K. Collys, T.F. Mets, D. Slop, M. Laska and D.L. Massart.

"Sensory detection of food odour in relation to dental status, gender and age". Gerodontology, 13 (1), 56-62 (1996).

Sensory detection tresholds are measured for aged individuals. A tendency for lower odour perception was observed for male partial denture wearers, especially those showing less oral hygiene.

   Y. Vander Heyden, D.L. Massart, Y. Zhu, J. Hoogmartens and J. De Beer.
"Ruggedness tests on the HPLC assay of the U.S. Pharmacopeia 23 for tetracycline.HCl : comparison of different columns in an interlaboratory approach". J. Pharm. Biomed. Appl., 14, 1313-1326 (1996).

Ruggedness tests were performed on the USP assay for tetracycline.HCl. Nine different columns, including C8 and C18 columns, were tested in three laboratories. Four factors were examined in half-fraction factorial designs, which were performed at three different column ages. The normalised effects of a factor on a response are found to be similar on the different columns and to be constant over time. In general C-8 columns were found to be preferred for this assay.

   A. de Juan, B. van den Bogaert, F. Cuesta Sánchez and D.L. Massart.
"Application of the needle algorithm for exploratory analysis and resolution of HPLC-DAD data". Chem. Intell. Lab. Syst., 33, 133-145 (1996).

The needle algorithm is based on a uniqueness test in target factor analysis. It is applied to simulated HPLC-DAD data sets and yields information on the number of analytes in the system, the presence of minor compounds and the location of peak maxima. How to use this information as input in ALS (Alternating least squares) is also discussed.

J.O. De Beer, C.V. Vandenbroucke, D.L. Massart and B.M. De Spiegeleer.

"Half-fraction and full factorial designs versus central composite design for retention modelling in reversed phase ion-pair liquid chromatography". J. Pharm. Biomed. Anal., 14, 525-541 (1996).

Half-fractional and full fractional designs are applied for the development of a separation of substances from a cough syrup. This separation was studied earlier with the central composite design. It is concluded that the full factorial design enables the selection of suitable LC parameter combinations.

   B. Walczak, B. van den Bogaert and D.L. Massart.
"Application of wavelet packet transform in pattern recognition of NIR data". Anal. Chem., 68, 1742-1747 (1996).

The wavelet packet transform is studied as a tool for improving pattern recognition based on near-infrared spectra. Application to the preprocessing of the spectra improves the classification when compared to using either the standard normal variate method or no pretreatment at all.

   W. Penninckx, C. Hartmann, D.L. Massart and J. Smeyers-Verbeke.
"Validation of the calibration procedure in atomic absorption spectrometric methods". J. Analyt. Atomic Spectrometry, 11, 237-246 (1996).

A general strategy for the validation of the calibration procedure in AAS is described. Therefore different experimental designs and statistical tests to trace outliers, to examine the behaviour of the variance and to detect lack-of-fit were evaluated.

   J. Toft, F. Cuesta Sánchez, B. van den Bogaert, F.O. Libnau, and D.L. Massart.
"Resolution of overlapping mid-infrared spectra using simplisma and a 2nd order derivative approach". Vibrational Spectroscopy, 10, 125-138 (1996).

Matrices of several overlapping mid-infrared mixture spectra have been resolved into the pure compounds' spectra and concentration profiles. For a three-compound mixture system of organic solvents, SIMPLISMA qualitatively resolves the spectra but provides only semi-quantitative concentrations. Quantitative results are obtained by applying an additional iterative procedure. Applied to water spectra acquired in the temperature range 2-96°C, SIMPLISMA resolves two spectra. These spectra can be interpreted as corresponding to water structures constructed of H₂O molecules with a relatively high and low average number of H-bonds.

   E. Bouveresse and D.L. Massart.
"Standardization of near-infrared spectrometric instruments : a review". Vibrational Spectroscopy, 11, 3-15 (1996).

This review describes the chemometric methods required to apply a calibration model developed on one instrument on data obtained on another one and to correct for differences arising in time.

   Bouveresse and D.L. Massart.
"Improvement of the Piecewise Direct Standardisation procedure for the transfer of NIR spectra for multivariate calibration". Chemom. Intell. Lab. Syst., 32, 201-213 (1996)

Artefacts in the piecewise direct standardisation method can be due to inadequate selection of the subsets of samples and incorrect local rank. Methods to detect and prevent the artefacts are described.

   E. Bouveresse, C. Hartmann, D.L. Massart, I.R. Last and K.A. Prebble.
"Standardisation of near-infrared spectrometric instruments". Anal. Chem., 68, 982-990 (1996).

Two different approaches are studied for the standardization of near-infrared spectrometric instruments. The first is the univariate slope/bias correction of the predicted values. The second approach is the piece-wise direct standardization (PDS). Both standardization methods are applied to three different data sets, and the results obtained are compared. A diagnostic tool is developed in order to decide whether the simple slope/bias correction approach can be applied instead of PDS.

   P. Vankeerberghen, J. Smeyers-Verbeke and D.L. Massart.
"Decision support system for run suitability checking and explorative method validation in electrothermal atomic absorption spectrometry". J Anal. Atomic Spectrometry, 11, 149-158 (1996).

A general strategy for the validation of the calibration procedure in AAS is described. Therefore different experimental designs and statistical tests to trace outliers, to examine the behaviour of the variance and to detect lack-of-fit were evaluated.

M. Jimidar, Q. Yang, J. Smeyers-Verbeke and D.L. Massart.
"Method development and optimization for small ion capillary electrophoresis". TrAC, 15 (n°2), 91-102 (1996).

The selection of initial conditions and the optimization of selectivity are reviewed. After dealing with the separation mechanisms for anions and cations, migration modelling procedures (both empirical and theoretical) are discussed.

   W. Penninckx, J. Smeyers-Verbeke, P. Vankeerberghen and D.L. Massart.
"Selection of reference or test materials for the validation of atomic absorption food analysis methods". Anal. Chem., 68, 481-489 (1996).

A strategy to select reference or other test materials on which method validation experiments should be performed is developed for the AAS analysis of foodstuffs.

   F. Cuesta Sanchez, J. Toft, B. van den Bogaert and D.L. Massart.
"Orthogonal projection approach applied to peak purity assessment". Anal. Chem., 68, 79-85 (1996).

The orthogonal projection approach (OPA), a stepwise approach based on an orthogonalization algorithm, is proposed. The performance of OPA for the assessment of peak purity in HPLC-DAD is described and compared with that of SIMPLISMA. The occurrence of artifacts in both approaches under nonideal situations is discussed.

   W. Wu, B. Walczak, D.L. Massart, S. Heuerding, F. Erni, I.R. Last and K.A. Prebble.
"Artificial neural networks in classification of NIR spectral data : design of the training set". Chemom. Intell. Lab. Syst., 33, 35-46 (1996).

Artificial neural networks (NN) with back-error propagation were used for the classification with NIR spectra and applied to the classification of different strengths of drugs. The NN architecture was selected through a pruning method, and batching operation, adaptive learning rate and momentum were used to train the NN. The results demonstrate that selection methods based on Kennard-Stone and D-optimal designs are better than those based on the Kohonen self-organized mapping and on random selection methods and allow 100% correct classification for both recognition and prediction.

S. Bouckaert, D.L. Massart, B. Massart and J.P. Remon.

"Optimization of a granulation procedure for a hydrophilic matrix tablet using experimental design". Drug Development and Industrial Pharmacy, 22 (4), 321-327 (1996).

An experimental design was used in order to optimize a granulation procedure in a high-shear mixer for a hydrophilic matrix tablet formulation. The parameters tested were the amount of water, the amount of granulation liquid, and the massing time.

S. Boonkerd, M.R. Detaevernier, Y. Vander Heyden, J. Vindevogel and Y. Michotte;

"Determination of the enantiomeric purity of dexfenfluramine by capillary electrophoresis : use of a Plackett-Burman design for the optimization of the separation". Journal of Chromatography A, 736 (1996) 281-289.

A capillary electrophoresis method for the chiral separation of the anorectic drug dexfenfluramine was improved. A two-level Plackett-Burman design was reflected to examine the influence of five method variables at three levels. The effects on the resolution, the analysis time and the peak asymmetry were evaluated by means of effect plots. Optimal values were predicted and led to experimental conditions capable to determine the levo-rotatory enantiomeric purity of dexfenfluramine.

   D. Jouan-Rimbaud, M.S. Khots, D.L. Massart, I.R. Last and K.A. Prebble.
"Calibration line adjustment to facilitate the use of synthetic calibration samples in NIR-analysis of pharmaceutical production samples. Anal. Chim. Acta, 315, 257-266 (1995).

A method is proposed to build a model for the quantitative determination of pharmaceutical tablets, where the active ingredient concentration does not vary enough. Synthetic tablets with an extended concentration range are used to build the model, based on MLR and feature selection.

   P. Vankeerberghen, J. Smeyers-Verbeke and D.L. Massart.
"The slope ranking method for homoscedastic calibration lines". Analusis, 23, 247-249 (1995).

The slope ranking method (SRM) detects possible deviations from linearity or blank problems. It is based on the ranking of the slopes between each data point and the origin. The method requires proper estimation of the probabilities for the detection of non-linearity of homoscedastic calibration lines. This article presents further work on the estimation of the type I error by means the of Monte Carlo technique.

Q. Yang, K. Desmit and J. Smeyers-Verbeke.

"Determination of trace cadmium in 24 h diets by graphite furnace atomic absorption spectrometry" J. Agric. Food Chem., 43, 2652-2657 (1995).

An analytical procedure for the determination of trace Cd in 24-h diets with graphite furnace atomic absorption spectrometry is developed and validated. Closed-vessel microwave acid digestion was used for the sample dissolution. To evaluate the bias of the method, reference materials were analyzed, and the results are in good agreement with the certified values. The precision of the method was evaluated by the analysis of a total diet sample. Cd is determined in 28 total diet samples collected during 7 days at four locations in Belgium. The estimated daily Cd dietary intake in Belgium is 18.4 ± 6.0 µg, which is acceptable compared with the ADI (60 µg/day) established by the FAO/ WHO.

I. Vercruysse, D.L. Massart and A.G. Dupont.

"Increase in plasma propranolol caused by nicardipine is dependent on the delivery rate of propranolol" Eur. J. Clin. Pharmacol., 49, 121-125 (1995).

The influence of a single dose of nicardipine 30 mg on the pharmacokinetics and pharmacodynamics of propranolol 80 mg given as a conventional release formulation and as a slow release formulation was studied in two separate groups of 12 healthy volunteers. Nicardipine doubled the area under the curve and C_max of propranolol when given as a conventional formulation, but increased it only slightly when given as a slow release formulation. These results suggest that the interaction is mainly due to short-term haemodynamic effects of nicardipine leading to saturation of hepatic enzymes or functional shunting.

   W. Wu, B. Walczak, D.L. Massart, K.A. Prebble and I.R. Last.
"Spectral transformation and wavelength selection in near infrared spectra classification". Anal. Chim. Acta, 315, 243-255 (1995).

Near-infrared spectroscopy is used to discriminate between different dosage strengths of tablets in blister packs. Univariate and multivariate feature selection and three data transformation methods are studied. The second derivative appears to be the most effective transformation. Univariate feature selection methods enable the selection of two selective bands, the difference of which can directly separate all groups. The bipolar axis in the biplots of scores and loadings obtained with PCA and PLS also enable the selection of discriminating ratios of wavelengths.

   D. Jouan-Rimbaud, D.L. Massart, R. Leardi and O.E. De Noord.
"Genetic algorithms as a tool for wavelength selection in multivarite calibration". Anal. Chem., 67, 4295-4301 (1995).

A comparison of MLR with PLS regression is presented for the multivariate modelling of hydroxyl number in a certain polymer of a heterogeneous near-IR spectroscopic data set. The MLR model was performed by selecting the variables with a genetic algorithm. It was shown that the MLR and PLS solutions are very similar. They include the same number of variables, and the first variables in each model have similar, chemically understandable functions. It is concluded that both solutions are equivalent.

   C. Hartmann, J. Smeyers-Verbeke, W. Penninckx, Y. Vander Heyden, P. Vankeerberghen and D.L. Massart.
"Reappraisal of hypothesis testing for method validation : detection of systematic error by comparing the means of two methods or of two laboratories". Anal. Chem., 67, 4491-4499 (1995).

The usual way of testing the equivalence of two measurement procedures is analyzed and possible alternatives are discussed. It is concluded that for method validation purposes defining an interval of acceptable values and applying interval hypothesis tests is preferred over the usually applied point hypothesis tests.

M.I. Griep, T. Mets, A. Vercruysse, I. Cromphout, I. Ponjaert, J. Toft and D.L. Massart.

"Food odor thresholds in relation to age, nutritional and health status". J. Gerontology, 50A (n°6), B407-B414 (1995).

The effect of aging and health status on detection of food odors and interrelation with nutritional status are explored. Young and elderly subjects are studied. The detection treshold of two food odors, one with a trigeminal and one with a olfactory effect, are measured. Significant declining sensitivity was observed for both. A relation between the nutrition and odor perception in the elderly population exists.

   A. Garrido Frenich, D. Jouan-Rimbaud , D.L. Massart, S. Kuttatharmmakul, M. Martinez Galera and J.L. Martinez Vidal.
"Wavelength selection method for multicomponent spectrophotometric determination using partial least squares". The Analyst, 120, 2787-2792 (1995).

A method for eliminating unnecessary wavelengths is applied with the goal of achieving improved prediction ability in multicomponent determinations by UV/VIS spectrophotometry with partial least squares (PLS). The feature selection method is based on the regression coefficients of the closed form of the PLS model. The results presented suggest that wavelength selection improves the prediction ability of PLS method.

Qing Yang, C. Hartmann, J. Smeyers-Verbeke and D.L. Massart.

"Method development and validation for the determination of mineral elements in food and botanical materials by capillary electrophoresis". J. Chromatogr. A, 717, 415-425 (1995).

A CE method was developed and validated for the determination of K⁺, Na⁺, Ca²⁺, Mg²⁺, and Mn²⁺ in solid natural products.

   Y. Vander Heyden, C. Hartmann, D.L. Massart, L. Michel, P. Kiechle and F. Erni.
"Ruggedness tests for an HPLC assay : comparison of an evaluation at two and three levels by using two-level Plackett-Burman designs". Anal. Chim. Acta, 316, 15-26 (1995).

Robustness tests were performed on a HPLC method to separate by-products from the main drug substance. Two-level Plackett-Burman designs were executed to evaluate the factors at two and three levels. Dummy factor effects and appropriately executed replicate nominal experiments were found suitable to estimate the experimental variability of an effect. Testing at three levels gives more information about the response function as a function of the factor levels.

   P. Fernandes de Aguiar, B. Bourguignon, M.S. Khots, D.L. Massart and R. Phan-Tan-Luu.
"D-Optimal designs". Chemom. and Intel. Lab. Systems, 30, 199-210 (1995).

Tutorial article. The D-optimality concept can be applied to select a design when the classical symmetrical designs cannot be used, such as when the experimental region is not regular in shape, when the number of experiments chosen by a classical design is too large or when one wants to apply models that deviate from the usual first or second order ones. The D-optimality concept is developed. A few other criteria are also given that complement the information obtained by the D-criterion.

   F. Cuesta Sanchez, J. Toft, Kvalheim and D.L. Massart.
"Eigenstructure tracking analysis for assessment of peak purity in HPLC-DAD". Anal. Chim. Acta, 314, 131-139 (1995).

Eigenstructure tracking analysis is applied for the detection of an impurity under a chromatographic peak and guidelines for the interpretation of the plots are given. A new normalization of spectra is proposed to remove heteroscedastic noise. The results obtained are compared with the performance of fixed size window evolving factor analysis, the methods based on the Gram-Schmidt orthogonalization and SIMPLISMA.

P.F. de Aguiar and D.L. Massart.

"Experimental design." In : AIP Conference proceedings 330 - E.C.C.C.1 Computational Chemistry - F.E.C.S. Conference, Nancy, France, 1994. Ed.: F. Bernardi and J-L. Rivail, Am. Inst. of Physics, N.Y., pg. 702-716, 1995.

General tutorial article.

   Vander Heyden, K. Luypaert, C. Hartmann, D.L. Massart, J. Hoogmartens and J. De Beer.
"Ruggedness tests on the HPLC assay of the United States Pharmacopeia XXII for tetracycline hydrochloride. A comparison of experimental designs and statistical interpretations". Anal. Chim. Acta, 312, 245-262 (1995).

Robustness tests were performed on the high-performance liquid chromatographic method for the determination of tetracycline hydrochloride from the United States Pharmacopoeia XXII using Plackett-Burman and fractional factorial designs. Statistical interpretation criteria applying dummy factor effects or two-factor interactions were compared with alternatives found in the literature and with graphical interpretation methods (e.g. normal probability plots).

C. Hartmann, W. Penninckx, Y. Van der Heyden, P. Vankeerberghen, D.L. Massart and R.D. McDowall.

"Experience with chromatographic methods - Europe." In : Bio-International 2, Bioavailability, Bioequivalence, and Pharmacokinetic Studies; by : H.H. Blume and K.K. Midha, Medpharm, Stuttgart, pg. 331-346, 1995.

Suggestions about the experimental designs and the data evaluation are presented for the validation of chromatographic bioanalysis methods. The following topics are discussed: (i) the process to select and validate the calibration model, (ii) the experimental design for routine application, (iii) the determination of bias and precision within one laboratory, and (iv) the establishment and validation of a method in another laboratory.

   M. Jimidar, B. Bourguignon and D.L. Massart.
"Selectivity optimization after prediction of the migration behaviour of anions in capillary ion analysis in the presence of micelles". Anal. Chim. Acta, 310, 27-42 (1995).

A central composite design is used to optimise the separation of 10 inorganic ions. Two types of models were used. The first one was an empirical second order polynomial, the second one was based on a theoretical model of the partition process in the presence of micelles.

M. Jimidar, C. Hartmann, N. Cousement and D.L. Massart.

"Determination of nitrate and nitrite in vegetables by capillary electrophoresis with indirect detection." J. Chromatography A, 706, 479-492 (1995).

Nitrate and nitrite (and some other anions) were determined in vegetables by capillary electrophoresis (CE). The CE method is divided into two parts: a high-concentration-level method (for nitrate determination) and a low-concentration-level method (for nitrite determination). These CE methods were compared with a reference method. Parameters such as linearity, detection limit, quantification limit, precision and accuracy of the two techniques were investigated and compared. Both techniques resulted in acceptable linearity within their ranges. The detection limits of the CE methods were sufficiently low for the determination of the anions in vegetable samples. The precision and accuracy of the CE methods were comparable to those of the reference method.

   F. Cuesta Sanchez, J. Toft, B. Van den Bogaert, D.L. Massart, S.S. Dive and P. Hailey.
"Monitoring powder blending by NIR spectroscopy." Fresenius J. Anal. Chem., 352, 771-778 (1995).

A strategy is proposed for the monitoring of powder blending. Samples are taken throughout the blender vessel and scanned by diffuse reflectance spectroscopy in the NIR range. The standard deviation spectrum at each time is calculated and the mean standard deviation value is plotted as a function of time. This plot indicates the time at which the blend is most homogeneous. Individual spectra can be compared with the "mixture" spectrum, which is an approximation of the spectrum of a true homogeneous sample. As an alternative approach to monitoring the blending the use of SIMPLISMA is recommended.

   M.I. Griep, I.N. Wakeling, P. Vankeerberghen and D.L. Massart.
"Comparison of semirobust and robust partial least squares procedures." Chemom. and Intell. Lab. Syst., 29, 37-50 (1995).

Three different robust partial least squares (PLS) alternatives are compared in order to reduce the effect of outlying data. Although iteratively reweighted least squares, based on the biweight mean, offers efficiency in a variety of situations, in higher dimensions, there are still considerable deviations from the ordinary PLS solution without outliers.

Q. Yang, Y. Zhuang, J. Smeyers-Verbeke and D.L. Massart.

"Interpretation of migration behaviour of inorganic cations in capillary ion electrophoresis based on an equilibrium model." J. Chromatogr., 706, 503-515 (1995).

In capillary ion electrophoresis, the migration of inorganic cations in an aqueous-organic medium is influenced by complexation and interactions between the solutes and organic solvents. A migration model is proposed that describes the electrophoretic mobility of the cations in terms of absolute mobility, complex formation constants and the concentrations of complexing reagents and organic solvent. It was validated in a background electrolyte composed of imidazole, 2-hydroxyisobutyric acid (HIBA), 18-crown-6 and methanol. The parameters in this model were estimated experimentally and were in good agreement with the values in the literature.

   C.L. Karr, B. Weck, D.L. Massart, P. Vankeerberghen.
"Least median squares curve fitting using a genetic algorithm." Engng. Applic. Artif. Intell., 8 (n°2), 177-189 (1995).

Least median squares (LMS) curve fitting is a method of robust statistics that guards the process of data analysis from perturbations due to the presence of outliers. Genetic algorithms are search techniques that model the search that occurs in nature via genetics. This paper presents a procedure for utilizing genetic algorithms in an LMS approach to curve fitting. Several examples are provided from a number of application areas, thereby demonstrating the versatility of the genetic-algorithm-based LMS approach.

   E. Bouveresse, P. Dardenne and D.L. Massart.
"Modified algorithm for standardisation of near-infrared spectrometric instruments." Anal. Chem., 67, 1381-1389 (1995).

The transfer of calibration models between near-infrared spectrometric instruments with the algorithm proposed by Shenk and Westerhaus as a standardization method is studied. This algorithm does not yield optimal results when standardization samples are used that are not of the same nature as the samples to be analyzed. The use of locally weighted regression, which gives more weight to the standardization samples close to the same spectral intensity range where the samples to be predicted are located and less weight to the farest samples, seems to yield good results.

   P. Vankeerberghen, J. Smeyers-Verbeke, R. Leardi, C.L. Karr and D.L. Massart.
"Robust regression and outlier detection for non-linear models using genetic algorithms." Chemom. and Intell. Lab. Syst., 28, 73-87 (1995).

Robust regression based on the calculation of the least median of squared residuals (LMS) is robust to the presence of outliers and is used for outlier detection. The original LMS program only handles models which are linear in the parameters. A genetic algorithm can be used to obtain the LMS parameters for models that are non-linear in the parameters. In this work the feasibility of using genetic algorithms for LMS is demonstrated by means of curved analytical calibration and pharmacokinetic data contaminated with outliers.

   D. Jouan-Rimbaud, B. Walczak, D.L. Massart, I.R. Last, and K.A. Prebble.
"Comparison of multivariate methods based on latent vectors and methods based on wavelength selection for the analysis of NIR spectroscopic data." Anal. Chim. Acta, 304, 285-295 (1995).

PCR with selection of principal components instead of the usual top-down approach yields simpler and better models. As feature selection methods, selection of wavelengths correlated with concentration, with large covariance with concentration, with high loadings on the important principal components, and other methods are considered. The presented results suggests that feature selection can improve multivariate caiibration.

   W. Penninckx, P. Vankeerberghen, D.L. Massart and J. Smeyers-Verbeke.
"Knowledge-based computer system for the detection of matrix interferences in atomic absorption spectrometric methods." J. Anal. Atomic Spectrometry, 10, 207-214 (1995).

A knowledge-based computer system which assists with the detection of matrix interferences during the validation of atomic absorption spectrometry (AAS) is described. The system compares the slopes of a standard addition line and an aqueous calibration line. An experimental design is proposed which guarantees that the probability of not detecting important interferences (β-error) is acceptably low. To permit a detection of interferences by the comparison of curved calibration lines, a linearization is proposed.

   B. Walczak and D.L. Massart.
"Robust principal components regression as a detection tool for outliers." Chemom. and Intell. Lab. Syst., 27, 41-54 (1995).

Robust principal components regression procedure based on the ellipsoidal multivariate trimming (MVT) and the least median of squares (LMS) methods is proposed as an outlier detection tool. The performance of this approach was evaluated using simulated data randomly contaminated.