In Silico SAR Studies of HIV-1 Inhibitors

Abstract

Quantitative Structure Activity Relationships (QSAR or SAR) have helped scientists to establish mathematical relationships between molecular structures and their biological activities. In the present article, SAR studies have been carried out on 89 tetrahydroimidazo[4,5,1-jk][1,4]benzodiazepine (TIBO) derivatives using different classifiers, such as support vector machines, artificial neural networks, random forests, and decision trees. The goal is to propose classification models that will be able to classify TIBO compounds into two groups: high and low inhibitors of HIV-1 reverse transcriptase. Each molecular structure was encoded by 10 descriptors. To check the validity of the established models, all of them were subjected to various validation tests: internal validation, Y-randomization, and external validation. The established classification models have been successful. The correct classification rates reached 100% and 90% in the learning and test sets, respectively. Finally, molecular docking analysis was carried out to understand the interactions between reverse transcriptase enzyme and the TIBO compounds studied. Hydrophobic and hydrogen bond interactions led to the identification of active binding sites. The established models could help scientists to predict the inhibition activity of untested compounds or of novel molecules prior to their synthesis. Therefore, they could reduce the trial and error process in the design of human immunodeficiency virus (HIV) inhibitors.

(Q)SAR Models of HIV-1 Protein Inhibition by Drug-Like Compounds

“…Ligand-based drug design methods are widely used in finding [11] and optimization [12] of new anti-HIV agents. Many (Q)SAR studies have been dedicated to the analysis of structure-activity relationships for particular classes of compounds, such as tetrahydroimidazobenzodiazepines [13] and hydroxyethoxymethylphenylthiothymines [14]. The applicability domain of the local (Q)SAR models developed with training sets including molecules from a single chemical class is typically limited to this particular chemical class.…”

Computational Models Using Multiple Machine Learning Algorithms for Predicting Drug Hepatotoxicity with the DILIrank Dataset

“…Random forests (RF) are a widespread classification algorithm in a variety of fields, including QSAR [80,81]. They assemble a large number of decision trees with the help of a simple majority vote to resolve the most probable class for each data point.…”

Application of Support Vector Machines in Viral Biology

“…To reduce the number of descriptors Attributes selection was carried out using 'Best-First' as the search method in Waikato Environment for Knowledge Analysis (Weka) suite. SMO regression algorithm in the Weka suite was [17]. They successfully employed seven molecular descriptors characterizing hydrophobic, electronic, and topological aspects of the molecules and obtained excellent training and test accuracies.…”