A Random Forest Model for the Analysis of Chemical Descriptors for the Elucidation of HIV-1 Protease Protein-Ligand Interactions

A model for the classification of 70 HIV-1 protease crystal structure binding pockets to one of its complexed FDA approved protease inhibitors utilizing Random Forest has been developed. 456 chemical descriptors of the binding pocket of each crystal structure have been computed and are used to develop the classification model. Simulations were performed to determine the optimal Random Forest model parameters. An implicit feature relevance measure for the optimal model was analyzed using the Gini importance measure. The chemical descriptors most influential in classifying the binding pockets of HIV-1 protease to its complexed protease inhibitor were analyzed and interpreted in terms of the binding pocket structure and their protein-ligand interactions. The selected descriptors by the Random Forest model provides insight on the structure of HIV-1 protease which can be used to drive the drug discovery process to design novel HIV-1 protease inhibitors.