MIA-QSPR and effect of variable selection on the modeling of kinetic parameters related to activities of modified peptides against dengue type 2

Multivariate image analysis applied to quantitative structure–property relationship (MIA-QSPR) has shown to be a useful tool to model the biochemical properties of a series of drug-like compounds. However, drawing and alignment of two-dimensional structures (the images) are usually performed manually, resulting in a few imperfections. Selection of descriptors, which are in fact pixels of images, can minimize such effects; also, it enables selection of those variables which indeed explains the variance in the activities block. Therefore, in order to obtain more parsimonious, predictive models, interval PLS (iPLS), genetic algorithm (GA), and ordered predictors selection (OPS) were applied to select appropriate MIA descriptors to model kinetic constants, namely substrate cleavage rate (kcat) and Michaelis (Km) constants, which correlate to the bioactivities of peptides against Dengue type 2 (DEN-2). The models built were used to predict kcat and Km of new proposed peptides, which are miscellany of substructures of the most promising peptides experimentally tested elsewhere.

► MIA-QSPR was used to model biochemical properties of modified peptides against DEN-2.
► Variable selection improves modeling power by removing spurious and collinear pixels.
► iPLS and GA-iPLS were the best methods for modeling kcat and Km, respectively.
► The best models were used to predict kcat and Km for novel drug-like congeners.