Comparative Molecular Binding Energy Analysis of HIV-1 Protease Inhibitors Using Genetic Algorithm-Based Partial Least Squares Method

Comparative molecular binding energy analysis (COMBINE) [1] is a helpful approach for estimation of binding affinity of congeneric ligands that bind to a common receptor. The essence of COMBINE is that the ligand-receptor interaction energies are decomposed into residue-based energy contributions, and then the partial least squares (PLS) analysis is applied to correlate energy features with biological activity. However, the predictive performance of PLS model drops with the increase of number of noisy variables. With regard to this problem genetic algorithm (GA) combined with PLS approach (GAPLS) [2] for feature selection has demonstrated the improvement on the prediction and interpretation of model. Therefore, the purpose of this paper is to derive a more accurate and more efficient GAPLS in COMBINE by introducing a number of successive refinements.