QSAR models and scaffold-based analysis of non-nucleoside HIV RT inhibitors

Abstract

A selection of 289 pyrimidine derivatives with anti-HIV RT activities as non-nucleoside HIV RT inhibitors (NNRTI) were studied. The associative neural network (ASNN) method was applied to develop a quantitative structure–activity relationship (QSAR) for anti-HIV RT activity. The calculated models were validated using the bagging approach. A consensus model with R2=0.87 and RMSE=0.5 was obtained from 10 individual models. Scaffold analysis and molecular docking of the compounds used in the QSAR model identified a potential chemical scaffold. The results showed that scaffold-based analysis of the QSAR model could be helpful in identifying potent scaffolds for further exploration than analyzing the overall model. Matched molecular pair analysis (MMPA) was applied in the QSAR model to characterize molecular transformations causing a significant change in the anti-HIV activity. The linear QSAR model was calculated to explore the structural features important for NNRTI activity. The results revealed that the activity of NNRT inhibitors is strongly dependent on their aromaticity and structural flexibility. The scaffold-based analysis of QSAR models with molecular docking and MMPA was found to be helpful in characterizing potential scaffolds for anti-HIV RT derivatives. The outcome of this study provides a deeper insight into the computer-aided scaffold-based design of novel molecules with HIV RT activities. It was also clearly shown that the consensus model's failure to correctly predict new chemical series could be due to the limitation of its applicability domain (AD). Redevelopment of models using new measurements can dramatically increase their AD and performance.