Combining PLS with GA-GP for QSAR

In this paper, a new algorithm, the partial least squares (PLS) improved by genetic algorithm–genetic programming (GA-GP) is applied to deal with functions for inner relationship in quantitative structure–activity relationship (QSAR). PLS is used to build a linear or nonlinear model between the principal components and its activity, and GA-GP is applied to regressions and equations. It develops PLS models to increase the range of PLS modeling. Using the inner relationship of polynomial function in this paper, a set of 79 inhibitors of HIV-1 reverse transcriptase, derivatives of a recently reported HIV-1-specific lead: 1-[(2-hydroxyethoxy) methyl]-6-(phenylthio) thymine (HEPT) was studied. The obtained QSAR model shows high predictive ability, rcv=0.900. It demonstrates that this method is useful.