Three-dimensional quantitative structure-activity relationship (3D-QSAR) models were developed for chromone derivatives against HIV-1 protease using molecular field analysis (MFA) with genetic partial least square algorithms (G/PLS). Three different alignment methods: field fit, pharmacophore-based, and receptor-based were used to derive three MFA models. All models produced good predictive ability with high cross-validated r2 (r2cv), conventional r2, and predictive r2 (r2pred) values. The receptor-based MFA showed the best statistical results with r2cv = 0.789, r2 = 0.886, and r2pred = 0.995. The result obtained from the receptor-based model was compared with the docking simulation of the most active compound 21 in this chromone series to the binding pocket of HIV-1 protease (PDB entry 1AJX). It was shown that the MFA model related well with the binding structure of the complex and can provide guidelines to design more potent HIV-1 protease inhibitors.
- chromone derivatives
- HIV-1 protease
- molecular field analysis (MFA)
- Machine Learning