Premium
Artificial neural network prediction of retention factors of some benzene derivatives and heterocyclic compounds in micellar electrokinetic chromatography
Author(s) -
Golmohammadi Hassan,
Fatemi Mohammed H.
Publication year - 2005
Publication title -
electrophoresis
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.666
H-Index - 158
eISSN - 1522-2683
pISSN - 0173-0835
DOI - 10.1002/elps.200500203
Subject(s) - quantitative structure–activity relationship , micellar electrokinetic chromatography , chemistry , lipophilicity , molecular descriptor , polarizability , artificial neural network , feature selection , kovats retention index , biological system , steric effects , linear regression , stepwise regression , benzene , benzene derivatives , computational chemistry , molecule , artificial intelligence , chromatography , capillary electrophoresis , machine learning , computer science , organic chemistry , stereochemistry , gas chromatography , biochemistry , chemical synthesis , in vitro , biology
A 5–4‐1 artificial neural network (ANN) was constructed and trained for prediction of the retention factors of some benzene derivatives and heterocyclic compounds in micellar electrokinetic chromatography (MEKC) based on quantitative structure–property relationship (QSPR). The inputs of this network are theoretically derived descriptors that were chosen by the stepwise variable selection techniques. These descriptors are: molecular surface area, maximum value of electron density on atom in molecule, path four connectivity index, average molecular weight, and sum of atomic polarizability which were selected by using stepwise multiple linear regression as a feature selection technique. The standard errors of training, test, and validation sets for the ANN model are 0.091, 0.119, and 0.114, respectively. Results obtained showed that nonlinear model can simulate the relationship between the structural descriptors and the retention factors of the molecules in data set accurately. Also the appearance of these descriptors in QSPR models reveals the role of electronic and steric interactions in solute retention in MEKC.