Identification of molecular fingerprints of phenylindole derivatives as cytotoxic agents: a multi-QSAR approach

Abstract

Phenylindole is reported to be an interesting scaffold having promising cytotoxic activities and can overcome the cancer drug resistance possibly via binding to the colchicine binding site of tubulin. In order to find out the molecular fingerprints for the better cytotoxic activity of phenylindole derivatives, multiple validated chemometric modeling approaches namely hologram QSAR (HQSAR), Bayesian classification model, and pharmacophore mapping analyses were applied into a dataset of 102 phenylindole derivatives. The final HQSAR model shows good statistical significance (Q2 = 0.760; R2Train = 0.868; R2Test = 0.660), and the best pharmacophore hypothesis has the highest regression coefficient value (r = 0.975) and the lowest RMS value of 0.679. Moreover, the Bayesian model is also statistically validated and robust to discriminate the cytotoxic and non-cytotoxic phenylindoles. These studies suggest that the amine group should be unsubstituted for retaining higher cytotoxicity. The pharmacophore mapping and Bayesian classification study suggest the importance of 2-phenyl group as a ring aromatic feature conducive to cytotoxicity. The steric and hydrophobic effect of long chain linear alkyl group has a positive influence on cytotoxicity as evidenced by the multi-QSAR study. Therefore, this multi-QSAR modeling reported here is beneficial in designing potential phenylindole cytotoxic agents in future.