Identification of new potential HIV-1 reverse transcriptase inhibitors by QSAR modeling and structure-based virtual screening

Abstract

Non-nucleoside reverse transcriptase inhibitors (NNRTIs) have gained a definitive place due to their unique antiviral potency, high specificity and low toxicity in antiretroviral combination therapies which are used to treat HIV. To design more specific HIV-1 inhibitors, 218 diverse non-nucleoside reverse transcriptase inhibitors with their EC₅₀ values were collected. Then, different types of molecular descriptors were calculated. Also, genetic algorithm (GA) and enhanced replacement methods (ERM) were used as the variable selection approaches to choose more relevant features. Based on selected descriptors, a classification support vector machine (SVM) model was constructed to categorize compounds into two groups of active and inactive ones. The most active compound in the set was docked and was used as the input to the Pharmit server to screen the Molport and PubChem libraries by constructing a structure-based pharmacophore model. Shape filters for the protein and ligand as well as Lipinski’s rule of five have been applied to filter out the output of virtual screening from pharmacophore search. Three hundred and thirty-four compounds were finally retrieved from the virtual screening and were fed to the previously constructed SVM model. Among them, the SVM model rendered seven active compounds and they were also analyzed by docking calculations and ADME/Tox parameters.

Keyword:

• Fingerprint
• SVM
• classification
• pharmacophore
• ADME/Tox

Disclosure statement

The authors have declared no conflict of interest.