Abstract

QSAR modeling, a powerful method for the computer - aided drug design, demands appropriate choice of molecular structure description. At present thousands descriptors of molecular structure are suggested in QSAR and QSPR approaches. The selection of a subset of the most relevant molecular descriptors, used as variables, is important step in model development. In this short review recently reported algorithms for variable subset selection procedure are considered. The scoring functions and some other useful guidelines are discussed.

Keywords: Variable subset selection, stepwise regression, stochastic search, partial least squares, QSAR prediction