4th EucheMs chemistry congress

thursday, 30-Aug 2012
s616
chem. Listy 106, s587–s1425 (2012)
Analytical chemistry Electrochemistry, Analysis, sample manipulation
Chemometrics – ii
o - 4 2 3
how to ACCeSS hidden inforMAtion in
ChroMAtoGrAPhiC dAtA
L. JohnSen 1
1 Faculty of Life Sciences, Department of Food Science,
Copenhagen, Denmark
In chromatographic data it is often seen that one peak reflect
more than one compound. Commercial software for multichannel
data handling (e.g. GC-MS) often offers the possibility
for performing deconvolution. However, the solutions given by
such software are often unreliable and in many cases there is little
or no possibility for the user to evaluate the quality of the resulting
model. An alternative to the manufacturing software is fitting of
Gaussian or Lorentzian models to the signal. However, the
solutions from such models are not unique. Another problem is
that the user must know how many compounds the model should
evaluate. Another approach is to use PARAlel FACtor analysis 2
(PARAFAC2), which has previously been shown to be a powerful
tool for resolution of overlapping peaks. [1, 2]
In the presentation it will be demonstrated how PARAFAC2
can help the analyst in the task with deconvolution of peaks.
New developments concerning automation will also be
presented. These developments enable the possibility for
non-chemometricians to make models with PARAFAC2 and will
result in a more unbiased result than if the models where
to be evaluated manually.
references:
1. J.M Amigo, M.J. Popielarz, R.M. Callejón, M.L. Morales
A.M. Troncoso, M.A. Petersen, T.B. Toldam-Andersen.
Journal of Chromatography A, 1217, 4422 (2010).
2. J.M. Amigo, T. Skov, J. Coello, S. Maspoch, R. Bro.
Trends in analytical Chemistry, 27, 714, (2008)
Chemometrics – iii
4 th EucheMs chemistry congress
o - 4 2 4
MuLtioBJeCtive exPeriMentAL oPtiMizAtion
L. A. SArABiA 1 , M. S. SánChez 1 , M. C. ortiz 2
1 University of Burgos, Mathematics and Computation, Burgos,
Spain
2 University of Burgos, Analytical Chemistry, Burgos, Spain
Most problems posed in an experimental framework have
several facets to be taken into account, starting with the
experimental factors and their influence in different analytical
responses. These different aspects tend to exhibit a conflicting
behaviour, the improvement of one of them results in deterioration
of some other(s).
Instead of weighting the different responses into a single one
to be optimized, the present work tackles the multicriteria
optimization from its vector nature to search for the
Pareto-optimal solutions, i.e. solutions that are optimal in at least
one of the criteria maintaining the rest in their best allowable
values. This multiresponse optimization is addressed from two
perspectives.
The most usual context: optimization refers to the searching
of experimental conditions to optimize several analytical
responses of interest. In general, this has to be approached from
an experimental perspective. Consequently, whether these
responses are individually or jointly optimized, the reliability of
the optimal solutions is dependent on a proper experimental
design.
For some experimental procedures, above all when there are
several experimental factors, the number of experiments in a
standard design may make it unaffordable. Hence, the other
perspective is the selection of the experimental design itself, based
on its characteristics. There are several criteria to measure the
quality of an experimental design (variance inflation factors,
values of the variance function and related to them the alphabetic
criteria). The search of a reduced design that maintains the
required quality is again a problem of multicriteria optimization.
By using analytical problems as guiding examples, Paretooptimal
solutions are computed for choosing suitable
experimental designs and for simultaneously optimizing several
analytical responses of interest. Besides, to study the information
in these optimal solutions a graphical way, an adapted version of
the parallel coordinates plot, is also shown.
Acknowledgments: Financial support through projects
CTQ2011-26022 and BU108A11-2
Keywords: Chemometrics; Experimental design; Analytical
methods; Gas chromatography;
AUGUst 26–30, 2012, PrAGUE, cZEcH rEPUbLIc