Application of random forest regression in the modelling of a newly synthesized compound as potent HIV-1 reverse transcriptase inhibitors

AIDS that is caused by HIV-1, threatens human life and spreads rapidly worldwide because of no effective vaccine. HIV-1 non-nucleoside reverse transcriptase inhibitors (NNRTIs), as an ingredient of highly active antiretroviral therapy (HAART), have played an important role in drug therapy [1, 2]. In this study, random forest (RF) model was developed for modeling and accurate prediction of anti HIV-1 inhibitory activities of azabenzene (pyrimidine, pyridine, pyridazine and triazine) derivatives as potent NNRTIs [3-7]. The chemical structure of each compound was converted to 29 molecular properties descriptors by Dragon software and 12 newly docking derived descriptors. Among total 41 descriptors, only 25 descriptors were selected after removing descriptors with pairwise correlation above 0.9 and used as significant descriptors, which relate the activity data to the molecular structures. The optimized RF model was developed and its prediction ability was evaluated by prediction of some compounds in the external (test) data set (n=10). The mean square error (MSE), mean absolute error (MAE) and correlation coefficient (R2) for test set were 0.0618, 0.2067 and 0.9546, respectively. The results obtained showed the superior prediction ability of the proposed model in the prediction of anti HIV-1 inhibitory activities.