QSAR modelling of anticancer activity of 17-picolyl and 17-picolinylidene androstene derivatives

Abstract

Cytotoxic activity (IC50) of a series of eighteen 17-picolyl and 17- picolinylidene androstene derivatives towards human prostate cancer cell line (PC-3) was predicted using artificial neural network approach. Prediction was based on in silico molecular descriptors selected applying stepwise regression and partial least squares methods. Two excellent neural networks were established. They can be used for prediction of cytotoxic activity of structurally similar compounds and can give necessary guidelines for synthesis of new anticancer agents.