Pixel selection by successive projections algorithm method in multivariate image analysis for a QSAR study of antimicrobial activity for cephalosporins and design new cephalosporins

Thirty-one Cephalosporin compounds were modeled using the multivariate image analysis and applied to the quantitative structure activity relationship (MIA-QSAR) approach. The acid dissociation constants (pKa) of cephalosporins play a fundamental role in the mechanism of activity of cephalosporins. The antimicrobial activity of cephalosporins was related to their first pKa by different models. Bidimensional molecular structures (images) were used to calculate some pixel descriptors. The selection of pixels by successive projections algorithm (SPA) was done to achieve simple MIA-QSAR models; based on a small subset of pixels. In the present study, we evaluated the performance of pixel selection technique using SPA for partial least squares (PLS) model. The obtained model showed nice prediction ability with root mean square error of prediction (RMSEP) values of 0.402, 0.315, and 0.160 for principal component regression (PCR), PLS and SPA-PLS models respectively. Among the three methods, SPA-PLS was potentially useful in predicting the pKa of cephalosporins. The study showed the maximum structural efficacy is on pKa in a, b and c regions.