Please login first
SAR Studies for the in-silico Prediction of HIV-1 Inhibitors
* 1 , 2 , 2 , 3 , 4 , 5 , 1
1  Department of Chemistry, Faculty of Science Semlalia BP 2390 Marrakech, Morocco
2  School of Health Sciences, University of KwaZulu-Natal, Westville, Durban 4000, South Africa
3  Sofia University "ST.KLIMENT OHRIDSKI" Faculty of Chemistry and Pharmacy, 1 James Bourchier Avenue 1164 Sofia, Bulgaria.
4  Ecole Nationale Supérieure d'Ingénieurs (E.N.S.I.) I. S. M. R. A., LCMT, UMR CNRS n° 6507, 6 boulevard Maréchal Juin, 14050 Caen France
5  Institute of Chemistry, Faculty of Natural Sciences and Mathematics, Ss. Cyril and Methodius University, Skopje, Macedonia

Published: 01 November 2017 by MDPI in 3rd International Electronic Conference on Medicinal Chemistry session ECMC-3

Tetrahydroimidazo[4,5,1jk][1,4]benzodiazepines (TIBO), as non-nucleoside analogues, constitute potent inhibitors of HIV-1 reverse transcriptase. In the present study, classification structure-activity relationship (SAR) models are developed to distinguish between high and low anti-HIV-1 inhibitors in this class of compounds. Different classifiers, such as support vector machines, artificial neural networks, random forests and decision trees have been established by using ten molecular descriptors. All models were validated using several strategies: internal validation, Y-randomization, and external validation. The correct classification rate ranges from 97% to 100% and from 70% to 90% for the training and test sets, respectively. A comparison between all methods was done in order to evaluate their performances. The contribution of each descriptor was evaluated to understand the forces governing the activity of this class of compounds.

Keywords: structure activity relationship; TIBO; HIV inhibitors; support vector machines; decision trees; random forests and artificial neural networks