Machine learning approaches in designing anti-HIV nitroimidazoles: 2D/3D QSAR, kNN-MFA, docking, dynamics, PCA analysis and MMGBSA studies

Abstract

In this study, newly synthesized 20 nitroimidazole derivatives were subjected to 2D and 3D Quantitative Structure-Activity Relationship (QSAR) study to investigate their anti-HIV activity against both ROD and IIIB strains. Later, proposed hypothesis was virtually proved by further in-silico studies. In statistically significant 2D-QSAR models r² values for IIIB strains 0.9241 and for ROD strains 0.9412 with corresponding q² values of 0.7706 and 0.8299, were obtained, respectively. Different models were constructed using three different kNN-MFA 3D QSAR approaches such as SW-FB, SA score, and GA. By using the generated hypothesis, newer analogues of nitroimidazole derivatives was designed and molecular modelling studies were conducted to prove the hypothesis. The three molecules were displayed the good docking scores compared to the reference molecule. The stabilities of docked complexes were analyzed by MD simulations and MMGB/SA calculations. These results offer insightful design guidance for novel anti-HIV compounds synthesis and suggest interesting directions for future pharmaceutical research.