Abstract
Quantitative structure-activity relationship (QSAR) analysis has been carried out with a series of 107 anti-HIV HEPT compounds with antiviral activity, which was performed by chemometrics methods. Bi-dimensional images were used to calculate some pixels and multivariate image analysis was applied to QSAR modelling of the anti-HIV potential of HEPT analogues by means of multivariate calibration, such as principal component regression (PCR) and partial least squares (PLS). In this paper, we investigated the effect of pixel selection by application of genetic algorithms (GAs) for the PLS model. GAs is very useful in the variable selection in modelling and calibration because of the strong effect of the relationship between presence/absence of variables in a calibration model and the prediction ability of the model itself. The subset of pixels, which resulted in the low prediction error, was selected by genetic algorithms. The resulted GA-PLS model had a high statistical quality (RMSEP = 0.0423 and R(2) = 0.9412) in comparison with PCR (RMSEP = 0.4559, R(2) = 0.7929) and PLS (RMSEP = 0.3275 and R(2) = 0.0.8427) for predicting the activity of the compounds. Because of high correlation between values of predicted and experimental activities, MIA-QSAR proved to be a highly predictive approach.