Prediction of antibacterial activity of pleuromutilin derivatives by genetic algorithm–multiple linear regression (GA–MLR)

Abstract

Use of quantitative structure–activity relationships for prediction of the antibacterial activity of pleuromutilin derivatives was studied. A suitable set of molecular descriptors was calculated and the important descriptors were selected by using the variable selections of stepwise multiple linear regression and genetic algorithm. Principal-components analysis was used to select the training set. The models were validated by use of leave-one-out (LOO) cross-validation, external test set, and the Y-randomization test. Comparison of the results obtained revealed the superiority of the genetic algorithm over the stepwise multiple regression method for feature selection. One genetic algorithm–multiple linear regression (GA–MLR) model with six selected descriptors was obtained. The root mean square errors of the training and test sets for the GA–MLR model were calculated to be 0.423 and 0.523, and the correlation coefficients were 0.839 and 0.807. The statistical parameter of LOO cross validation test correlation coefficients on the GA–MLR model was 0.760. The predictive ability of the model was satisfactory and it can be used for designing similar groups of antibacterial compounds.

Graphical abstract