Abstrait

A simple approach to the prediction of electrophoretic mobilities of some peptides

Rahmatollah Rajabzadeh, Maziar Ahmadi Golsefidi, Gozal Inchebroni

In this study, based on molecular structure descriptors, by the use of partial least squares analysis, a good prediction quantitative structureproperty relationship for the mobility of some Peptideswas obtained. Five classes ofmolecular descriptors including topological, geometrical, functional group, empirical and properties descriptors were generated for each peptide. Constant and near constant variables exclude from descriptors and then descriptors with pair correlations 0.95 exclude, the total remaining descriptors were 72 and we use these descriptors for Genetic Algorithm (GA) variable selection.Aseries of 12 peptides were selected randomly to check the prediction ability of the obtained model. The RMSEP was 0.492 for selected model.