|
|
Research Topics
The Analytical Chemistry research group within the department
Analytical Chemistry and Pharmaceutical Technology (FABI) is performing
research in Applied Chemometrics and in Separation Science-related
topics. Earlier, two groups could be distinguished within FABI,
the Chemometrics and Metrology group headed by profs. D.L. Massart
and J. Smeyers-Verbeke, and the Separation Group by Y. Vander
Heyden. Unfortunately, Prof. Massart passed away in 2005. At the
moment the distinction between both groups is less pronounced
since Y. Vander Heyden heads, both together with prof. J. Smeyers-Verbeke
and prof.D. Mangelings. However, it is our goal to continue research
in both domains.
The goals of the Separation group are to apply chemometrics in
separation science on the one hand, and to evaluate new developments
in separation science on the other. At the moment we focus on
certain areas of separation science, such as chiral separations,
fingerprint development and the application of chemometrics in
chromatographic processes and data analysis.
In the following an overview is given of the ongoing or recently
performed research topics within FABI
Topics
Some more information on the different topics
Chiral separations and the development of a knowledge based system (D. Mangelings, H. Ates, A. Younes, A. Hendrickx, K. De Klerck)
The interest in the development of technologies to resolve isomeric
drugs is continually increasing due to, among others, more and
more stringent regulatory requirements. Chromatographic and electrophoretic
techniques are the most popular in this field. Chiral discrimination
mechanisms are complex and therefore the enantioselectivity of
a chiral selector towards a molecule is up till now hardly predictable.
Furthermore, over one hundred types of chiral chromatographic
columns and many chiral selectors are available. On the other
hand, the time dedicated to method development in industry is
always decreasing. Therefore, a knowledge-based system (KBS) for
chiral separations would be valuable to speed up method development.
The KBS developed earlier in our laboratory combines chemical
and experimental design knowledge. Multi-criteria methods, such
as Pareto optimality and Derringer functions are used to select
the best experiment after performance of an experimental design
and when more than one response is regarded. The system includes
3 modules: technique selection, screening and optimisation. Five
techniques, Reversed-Phase Liquid Chromatography (RPLC), Normal-Phase
Liquid Chromatography (NPLC), Polar Organic Solvent Chromatography
(POSC), Supercritical Fluid Chromatography (SFC) and Capillary
Electrophoresis (CE), are included. The development of the separation
strategies and of the KBS was done in co-operation with Sanofi-Aventis,
Toulouse, France. Method development can be followed by robustness
testing using the in-house developed software "Software for Robustness
testing" (SRT), created in order to simplify the process of setting
up and interpreting robustness tests.
At the moment, H. Ates evaluates the use of newly developed chiral
stationary phases (polysaccharide phases with bonded selectors,
polysaccharide phases based on chlorinated selectors) for their
applicability in chiral separations in POSC.
A. Younes uses the same chiral stationary phases (CSP), but in
normal-phase liquid chromatography and reversed-phase liquid chromatography.
It is the intention that, if the new stationary phases show adequate
enantioselectivity towards a large number of structurally diverse
compounds, that the existing strategies of the KBS are updated,
or that new separation strategies are defined when needed. This
will provide a possibility to update the existing strategies of
the KBS in terms of the increasing number of commercially available
chiral stationary phases.
Experiments in supercritical fluid chromatography (SFC) mode were
recently started by K. De Klerck. She will explore the CSP investigated
by H. Ates and A. Younes, include some recently commercialized
CSP in order to adapt also the earlier developed screening step
in this separation mode. She will also define a complete separation
strategy in SFC, including the optimization steps. Finally, she
will try to find the most suitable alternative technique to propose
in the decision trees when a given technique does not work. This
study will be performed by applying chemometric techniques to
find the most (dis)similar systems.
D. Mangelings examined the applicability of capillary electrochromatography
(CEC) in chiral separations. CEC is an analytical technique, combining
both high-performance liquid chromatography (HPLC) and capillary
electrophoresis (CE). It uses capillary columns filled with stationary
phase, like in HPLC, and an electrical field is applied to perform
separations, like in CE. The initial goal was to see if CEC was
a technique that allowed defining generic separation strategies,
applicable on a large number of compounds with chemical and structural
diversity. It was seen that polysaccharide CSP served best for
this purpose and a chiral separation strategy fulfilling the initial
goal could be defined on these stationary phases.
When using capillaries packed with particle-based stationary phase,
generic separation strategies could be defined. However, the technique
lacks robustness due to the fragility of the columns. Therefore,
efforts are being made to prepare capillary monolithic columns
for chiral separations, which might overcome the problems observed
with the particle-based columns. Here, the two same approaches
will be evaluated as with silica-based particles, i.e. the addition
of a chiral selector to the mobile phase, or its binding onto
the monolithic matrix. It is again the intention to define generic
separation strategies. Another type of stationary phases that
will be evaluated are the particle-glued stationary phases, where
the particles of classical stationary phases are glued together
through a sol gel procedure, and which also eliminate the need
for frits.
In this context, classical chiral particle-based stationary phases
are tested by A. Hendrickx for their enantioselectivity. Actually,
she uses the same CSP as in HPLC and SFC modes. When an improved
enantioselectivity is observed also in CEC mode with these phases,
the existing separation strategy will be updated. Then, the most
enantioselective phases will be converted into particle-glued
CSP and again enantioselectivity will be evaluated. This will
allow concluding to which extent the glueing process has an influence
on the separation performance of the CSP.
Back to top
Fingerprint development and data analysis (C. Tistaert, B. Dejaegher)
Chromatographic fingerprints are chromatograms, characteristic
for a given complex sample. The development and data analysis
of fingerprints is a research topic that is gaining increasing
interest in the pharmaceutical and biomedical analysis. Fingerprints
are, for example, developed to detect fraud in the synthesis of
drugs and food additives. They allow, for instance, to verify,
from the impurity profile, whether a molecule has been synthesized
using the patented synthesis route, or whether food additives,
like vanillin, come from a natural, microbiological or synthetic
origin. Such identification is of economical importance. Secondly,
fingerprints are also applied to analyze phytotherapeutic products,
in order to characterize herbal extracts with the purpose to avoid
switching and to result in a better quality control of herbs or
extracts. The complexity of the samples often makes it impossible
to separate, identify, and quantify all components. Therefore,
in the past, often only a selected number of components was determined,
which did not always allowed evaluating and judging the intrinsic
quality of the sample. For this purpose, the chromatographic fingerprint
technology, where the full chromatogram will be used for the extraction
of information, was developed. Since 1991, this technology was
introduced and accepted by the World Health Organization as an
identification and quality-evaluating technique for medicinal
herbs. Since 2000, the Chinese State Food and Drug Administration
demands fingerprints for the quality control of certain herbal
medicines in order to standardize the herbs and their preparations.
Fingerprints are also accepted by the USA Food and Drug Administration
as an alternative approach to identify a herb, and by the European
Agency for the Evaluation of Medicinal Products when the stability
of each individual component separately can not be determined.
Also Belgian companies that process herbal extracts in their preparations
use fingerprints as a quality criterion. Thirdly, also in drug
discovery, a screening strategy based on fingerprints could lead
to an important improvement. Many commercial drugs are based on
molecules discovered in natural organisms, such as plants or bio-organisms.
The number of drugs, only developed using drug design methodology,
is still rather limited. In view of the large number of plants
and bio-organisms with potential medicinal components, a fast
methodology is needed that allows to separate herbal and bio-extracts,
to evaluate certain activities (e.g. antioxidant, antimicrobial,
or cytotoxic activity), and to screen for components responsible
for the observed activity.
Fingerprints often contain many overlapping peaks. Therefore,
during fingerprint development it is a challenging task to select
conditions where a maximal number of components is separated.
Secondly, after fingerprint development, the useful information
should be extracted from the fingerprints, and linked either to
the identification and/or quality control of the sample, or to
its activity. For this purpose, multivariate data analysis techniques
are indispensable. The research considers three parts: (1) the
development of fingerprints, (2) the data pretreatment, and (3)
the extraction of relevant information.
Back to top
Process Analytical Technologies (S. Pieters,
B. Dejaegher)
This research topic tackles Process Analytical Technologies (PAT)
– related research, which is another recent research theme
within the pharmaceutical industries. The goal of PAT is to guarantee
the desired product quality by developing fully understood, robust,
continuously controlled and efficient production processes. Chemometrical
techniques can be applied to optimize the processes on the one
hand, and to extract the desired information out of the huge amount
of data on the other hand. To get full benefit out of on-line
PAT tools (e.g. NIR) to predict the quality of the product, robust
calibration methods need to be evaluated and need to be validated
afterwards. In this context, FABI focuses on finding the best
way to detect outliers, to select the best variable range and
to find a way dealing with the linearity problem of on-line NIR
methods.
Back to top
Separometrics: Contributions to faster development
and validation of pharmaceutical analytical separation methods
(B. Dejaegher)
Separometrics is the use of chemometric methods in separation
techniques. Research is performed on three subjects. In a first
research topic, the applicability of minimal and less frequently
used experimental designs in method development/optimization and
robustness testing is examined. Method development is often divided
into a screening phase where the most important factors are identified,
and an optimization phase where the most influencing factors are
further optimized. In the context of screening and robustness
testing, supersaturated designs were already examined. D-optimal
and Rechtschaffner designs will also be evaluated in this context
as alternatives to the classic fractional factorial and Plackett-Burman
designs. In the context of optimization, the applicability of
optimal and uniform designs as response surface designs will be
evaluated. In a second research topic, the development and multivariate
data analysis of fingerprints will be tackled. Fingerprints will
be developed according to a defined strategy. Subsequently, the
data will be pre-treated and the relevant information will be
extracted using multivariate data analysis techniques, such as
multivariate calibration methods. In a third research topic, Process
Analytical Technologies (PAT) – related research will be
performed.
Back to top
Development and validation of simple methods to assay drug formulations (D. Shewiyo, B. Dejaegher)
Experimental designs are also applied in method validation, more
specifically during robustness testing, e.g. in the context of
a Belgian-Tanzanian Bilateral Scholarship (Belgian Technical Cooperation
(BTC)) Program on the development and validation of simple methods
to assay drug formulations.
In a first study a normal-phase HPTLC method was developed for
the simultaneous analysis of sulfamethoxazole and trimethoprim
in co-trimoxazole tablets, and in a second study of lamivudine,
stavudine and nevirapine in fixed-dose combination tablets. Both
methods were validated and the linearity, precision, trueness,
and specificity were evaluated. A robustness test was also performed
using an 8-experiments Plackett-Burman design. The results from
the validation/robustness test were acceptable, and both methods
were found to be linear, precise, true, and robust.
Back to top
Comparison of feature selection and modeling
approaches in QSAR/QSPR (M. Goodarzi)
Mohammad’s work focuses on the comparison of different
feature selection and modeling methods in Quantitative Structure-Activity
Relationships (QSAR) and Quantitative Structure-Property Relationships
(QSPR) to predict either the biological activity or properties
of molecules. QSAR/QSPR are mathematic models to describe the
relationship between a biological activity or property and the
molecular descriptors of molecules. Two points are taken into
account: first, different available softwares are tested to find
the most stable conformer compared with x-ray crystallography
and docking information (protein-ligand interaction). When a good
and optimized structure is obtained, different types of descriptors
are calculated. Different feature selections techniques, not only
filter but also wrapper methods and especially swarm intelligences
are applied to find the most relevant descriptors from the pool.
On the other hand, different linear or nonlinear modeling methods
are applied to make a robust relationship between the selected
descriptors and property or activity. Therefore, these studies
can be useful in drug discovery and development and for the design
of new compounds with improved potency.
Back to top
Evaluation of fused-core columns for fingerprint development (G. Parewyck, D. Mangelings)
Greet’s project mainly focuses on the use of fused core
silica particles as stationary phases in HPLC. These stationary
phases use particles that have a solid core, surrounded by a porous
layer. Earlier, porous particles were mostly used as basis of
the stationary phases in HPLC. Hence, a substance can migrate
through the particles, which results in band broadening, an undesired
phenomenon because it decreases the global efficiency of the column.
Fused core particles eliminate the possibility of migration through
the particle center due to the solid core, and would theoretically
result in faster and more efficient analyses. It is claimed that
these stationary phases can compete with UHPLC, a technique that
uses ultra high pressures in combination with sub 2 µm particle-based
stationary phases to achieve these properties. This project will
evaluate whether the predicted behavior of fused core particle-based
stationary phases is also seen in practice. Fingerprint chromatograms
will be developed on these new stationary phases, on monolithic
phases (which also allow fast analyses) and also on classical
phases. A comparison between the results will reveal whether the
fused core particles show a behavior which is more related to
HPLC on monolithic phases or to classical particles. Occasionally,
a comparison with UHPLC on sub 2 µm particles can also be
made.
Back to top
Formerly Studied Topics
Some more information on the different topics
Use of analytical separation techniques to elucidate viral interactions at molecular level (I. Oita)
In cooperation with the department of Pharmaceutical Biotechnology and Molecular Biology (Prof. B. Rombaut, H. Halewijck) of the VUB, a project was started that aims to link analytical separation techniques to virological research. Using a cell-free system for the production of virus material and capillary electrophoresis as separation technique, it is the intention to study some unresolved issues regarding the structure and replication of picornaviruses, such as the function of some viral proteins, the mechanism of initiation of protein- and RNA-synthesis, and the different steps in the morphogenesis. It should also enable i) to detect subviral particles formed during the replication cycle of picornaviruses (morphogenesis), ii) to study the interactions between the viral genome and subviral particles, which leads to the formation of new virions, iii) to study interactions between viral RNA and cellular proteins, iv) to consider interactions between viral and cellular proteins, v) to monitor interactions between viral RNA and viral proteins with particular attention for the determination of the stoichiometry and affinity of, and finally vi) to find and identify new targets for potential antiviral drugs.
Back to top
Monolithic columns for capillary electrochromatography (I. Tanret)
Indiana Tanret is working on the implementation of polymeric monolithic stationary phases in columns for (pressurised) capillary electrochromatography and flat-rectangular channel column chromatography within the framework of the IWT-SBO Nextchrom project. These monolithic phases are an alternative for the more commonly used particulate phases, which have many disadvantages. The focus of the work was testing the applicability of these in-house synthesized stationary phases in drug analysis.
Back to top
Orthogonal chromatographic systems in pharmaceutical analysis: selection, applications, method development and data-analysis (M. Dumarey)
The aim of the project was to select chromatographic systems
that are as orthogonal or as dissimilar as possible, i.e. which
have the most different selectivity. With a chromatographic system
is meant the combination of (i) a stationary phase, (ii) a mobile
phase buffer at a given pH and (iii) an organic modifier type.
In this context the applicability and selectivity of new stationary
phases, as for instance polystyrene-divinyl copolymer, zirconia
based columns (e.g. polybutadiene coated, polystyrene coated,
graphitized carbon-clad zirconia), polymeric monolithic columns,
fluorinated reversed-phases is examined. This means that their
selectivity is compared among each other as well as relative to
traditional C8 or C18 reversed-phase columns.
To separate a mixture
with an unknown composition, for instance a new drug and its impurities,
it is useful to perform a screening on a set of chromatographic
systems with varying selectivities. This results in complementary
information about the analyzed sample.
One chromatographic system,
on which the best separation was obtained, is retained for further
method optimization. A methodology will be defined to optimize
sequentially the pH and composition of the mobile phase, the temperature
and the gradient slope.
A second application for which orthogonal
chromatographic systems might be advantageous, is fingerprinting.
Fingerprints are mainly used to obtain information about the identity
or the quality of phyto-therapeutical products with a complex
composition. In this project it will be studied whether combining
the data of two orthogonal fingerprints gives better classification
and multivariate calibration results.
Back to top
Multivariate calibration (X. Capron)
- Multivariate calibration using different models, mainly PCR
and PLS models, more particularly the updating of these models,
i.e. how to improve the model when it has to deal with new sources
of variance which were not taken into account during the initial
calibration step. Various domains of chemometrics were used, such
as model selection and validation, information complexity, variable
selection, correction of signal, filtering with methods such as
the Orthogonal Signal Correction (OSC) or the Net Analyte Signal
(NAS).
- WineDB European project, which aims identifying
the country of origin of wines. This work is mainly based on discrimination,
classification and feature selection. For this work, mainly techniques
such as CART and UVE-PLS are used. Some PCA and PP analysis has
also been performed.
- Estimation of the optimal complexity of
a PLS regression model in cases where atypical samples are present
in the calibration data, which prevent from using cross-validation
methods. Atypical samples are defined as samples fitting to the
model in calibration, but behaving as outliers in prediction.
Back to top
Chemometrical exploration of environmental data
(I. Stanimirova)
The work focused on the chemometrical exploration of environmental
data. In this context, some new chemometrical approaches were
introduced in environmental science as alternatives to the already
existing and their usefulness was compared for interpretation
and visualization of environmental data. In order to derive a
strategy for exploration of multi-dimensional environmental data:
- Projection (as PCA, projection pursuit (PP) with different projection
indices) and clustering techniques (as hierarchical, density-based
approaches, self-organizing maps, neural gas) were explored.
- A comparative study of two robust PCA algorithms based on PP
was carried out on different data sets containing various numbers
of objects and variables as well as different observation structures.
- The performance of a method for three-way analysis called STATIS
(Structuration des Tableaux A Trois Indices de la Statistique)
was compared with Tucker3 and PARAFAC2, two more commonly used
methods.
- A four-way environmental data set, where missing elements were
present, was explored by using the PARAFAC and Tucker methods.
- The multiple factor analysis (MFA) approach was introduced as
an alternative to Procrustes and canonical correlation analyses
methods for dealing with data with different blocks of variables,
possibly of different type. Furthermore, it was shown that the
MFA method can be extended such that a variable selection can
be performed by adapting methods applied in PCA for this purpose.
The above-mentioned methods were not only used for interpretation
and visualization of environmental data sets. For instance, the
use of STATIS was explored for monitoring batch processes (in
collaboration with S. Gourvenec). Furthermore, the usefulness
of PP combined with Multivariate Regression Trees (MRT) was shown
for explaining the presence of groups in analytical data in terms
of original variables.
Back to top
Alternative methods for multivariate calibration
(L. Jin)
The major subject is to propose some alternatives to PCR/PLS
for multivariate calibration. In a first stage of the study, the
Law of Mixtures (LM) and the Delaunay Triangulation (DT) methods
are proposed as new local multivariate calibration methods. In
the second stage, updating multivariate calibration with the DT
method is investigated.
The proposed LM/DT method can be considered as a Nearest Neighbors
method or as what is sometimes called a topological method. In
these methods, a mesh of simplexes in the calibration data set
is constructed firstly. The result for an unknown sample is then
considered as the weighted average of the y-values of the calibration
samples that constitute the simplex containing or close to the
unknown sample. In all cases studied, the LM/DT methods give at
least similar results as PCR/PLS.
Back to top
Micellar liquid chromatography in the screening
for potential drugs (A. Detroyer)
When screening for drug candidates it is essential to know the
extent to which a substance passes biological membranes, because
its pharmacological effect depends upon it. In-vitro models of
bio-membranes (e.g. Caco-2), although good predictors, do not
allow high-throughput screening. Therefore attention is paid to
other, e.g. chromatographic, methods. Investigations on classical
RPLC systems have been made to correlate retention with log P.
This log P is thought to be a parameter representing the main
transport mechanism when passing a bio-membrane. Although correlation
between log P and the bio-membrane permeability is possible, it
is thought that newer chromatographic systems, with more interfacial
characteristics, will better represent permeability. This is confirmed
using, for instance, immobilised artificial membrane (IAM) columns.
However, micellar liquid chromatography (MLC) seemed to be a cheaper
alternative. The aim of this project was to make MLC operational
as an alternative methodology when trying to predict the biological
activity of candidate drugs and to evaluate it in a QSRR (Quantitative
Structure Retention Relationship) context. The MLC retention parameter
was compared with other bio-membrane passage determining parameters
like the log P, the IAM predictor and the Caco-2 predictor. Part
of the experiments in this project were performed in co-operation
with and at the Janssen Pharmaceutica site in Beerse. Good agreement
in results was found and the technique could be considered a fast
alternative for the other methods because MLC on monolithic columns
could be applied which makes the measurements very fast.
Back to top
Combination of chromatographic and molecular
descriptors to predict the membrane passage of drugs (E. Deconinck)
A major problem in drug discovery is that molecules, found to
be potentially useful as drugs, fail in a later stage of the drug
development due to non proper ADME-Tox-properties. Therefore methods
are necessary that allow screening for these properties in the
first stages of the drug development.
The aim of this project is to combine in-silico (theoretical molecular
descriptors) and fast in vitro (chromatographic descriptors, e.g.
MLC retention) parameters in models to predict the intestinal
absorption or the blood-brain barrier passage of candidate drug
molecules. This project was a kind of continuation of the previous
one on the use of MLC in this context. In this project the use
of theoretical descriptors as well as chromatographic descriptors
were evaluated in a QSAR-context. The chromatographic descriptors
consist of the retention factors measured on different HPLC systems.
A set of orthogonal reversed phase-HPLC systems as well as IAM-
and MLC-systems were considered. Modelling was performed using
relatively new techniques in the QSAR-domain, like CART and MARS.
The use of CART (Classification And Regression Trees), MARS (Multivariate
Adaptive Regression Splines) and two step-MARS (TMARS) were already
evaluated for the prediction of gastro-intestinal absorption of
drugs. The improvement of the obtained models by adding the measured
chromatographic descriptors were evaluated, and also other chemometric
techniques were evaluated for their use in the prediction of gastro-intestinal
absorption: Boosted Classification Trees and different two-step
modeling techniques.
The use of these techniques was also evaluated for the prediction
of other molecular or sample properties, as e.g. blood-brain barrier
passage, the anti-oxidative capacity of green tea and the inhibitory
effect of certain drugs on enzyme activity.
Back to top
Regression trees and MARS in QSRR and in the
selection of orthogonal systems (R. Put)
This research consisted of two subjects, both dealing with chemometrical
applications on chromatographic data. The major topic was quantitative
structure-retention relationships (QSRR). QSRR's are models that
predict the chromatographic retention of solutes, based on their
physico-chemical or structural properties, which are encoded in
molecular descriptors. In order to build meaningful models, appropriate
descriptors should be selected from a wide range of available.
The methods studied in the QSRR context were new in this research
area, such as classification and regression trees (CART) and multivariate
adaptive regression splines (MARS).
The second topic dealt with the selection of the most orthogonal
chromatographic systems. In this context the use of (univariate)
CART-bushes and auto-associative multivariate regression trees
(AAMRT) was studied. QSRR's were also studied a) for a data set
of peptides measured on a series of chromatographic systems and
b) for a relatively large set of drugs on orthogonal chromatographic
systems. In these contexts, two-step MARS, MARS and MLR were compared.
The use of Genetic Algorithms for the selection of proper molecular
descriptors was studied too.
Back to top
Prediction of enantioselectivity (S. Caetano)
The aim of this research was to predict the enantioselectivity
of given systems towards the chiral separation of compounds, i.e.,
to build models that are able to indicate whether a separation
between two enantiomers is possible or not. At present time, if
someone wishes to perform a chiral separation of some substance
one must seek the suitable chiral stationary phase (CSP) by trial
and error, making the selection time consuming and expensive.
Therefore, the possibility to predict whether or not a certain
CSP is able to perform a chiral separation would be of great benefit.
To reach this goal several chemometrical techniques were being
used, such as Classification and Regression Trees (CART), Partial
Least Squares (PLS), Stepwise Multiple Linear Regression (stepwise-MLR)
and Multivariate Adaptive Regression Splines (MARS). These methodologies
were used to build models that are able to correlate the molecular
descriptors (occasionally able to distinguish between enantiomers,
such as chirality codes) with the enantiomeric selectivity. Different
responses to describe enantioselectivity like the separation factor
and the resolution were used to build the models.
Back to top
Fast chromatography (A.M. Van Nederkassel)
In this research topic it was evaluated what are the possibilities
of very fast chromatographic methods (with analysis times of e.g.
less than one minute). Monolithic columns were used for this purpose.
Such fast chromatograms with a short run time can be treated as
if it were spectra. This means that the chemometric techniques
used in spectroscopy can be applied. For instance multivariate
calibration, in which the complete chromatogram is used to model
and predict a given property, could be used. This was done, for
instance, to predict the antioxidant capacity of green tea extracts.
These multivariate analyses also required pretreatment of the
chromatograms, e.g. to align a given peak in the different chromatograms.
To estimate concentrations of individual compounds overlapping
with others in the chromatogram, peak resolution methods are required.
In the above context, the following applications were investigated:
- Method development and evaluation of the transfer - and acceleration
- of some separations, developed on classical HPLC columns, to
monolithic columns. The robustness of the transferred separations
was also examined.
- Development of a fast separation method for green tea extracts
on a monolithic column and estimation of the antioxidant capacity
of the extracts from their chromatogram building a multivariate
calibration model. This is an application where the complete chromatogram
is used to estimate a global property. The following techniques
are used: Correlation Optimized Warping to align the chromatograms,
robust PCA (R-PCA) to detect leverage objects and PLS and Uninformative
Variable Elimination-PLS to build a model.
- Development of a fingerprint chromatogram for the Ginkgo Biloba
extract on a monolithic stationary phase, in the context of determining
the quality of herbal medicines. A screening design was used to
optimize the operating parameters of the ELS detector.
- Comparison of the performance of three aligning methods for
different data sets: Correlation Optimized Warping (COW), Parametric
Time Warping (PTW) and Semi-Parametric Time Warping (STW).
Back to top
|
|