Ligand similarity guided receptor selection enhances docking accuracy and recall for non-nucleoside HIV reverse transcriptase inhibitors

• Type of Material: Article
• Authors: Richard A. Stanton, Emory University; James Nettles, Emory University; Raymond Schinazi, Emory University
• Language: English
• Date: 2015-11-01
• Publisher: SPRINGER
• Publication Version: Author Accepted Manuscript (After Peer Review)
• Copyright Statement: 2015
• Final Published Version (URL): http://dx.doi.org/10.1007/s00894-015-2826-7
• Title of Journal or Parent Work: JOURNAL OF MOLECULAR MODELING
• Volume: 21
• Issue: 11
• Start Page: 282
• End Page: 282
• Grant/Funding Information: This work was supported in part by Public Health Service grants 5P30-AI-50409 (CFAR), and by the Department of Veterans Affairs. Pipeline Pilot and Discovery Studio software were received as Academic Achievement awards from Accelrys Corporation.
• Keywords: Computer Science, Interdisciplinary Applications; Science & Technology; MECHANISMS; Physical Sciences; COMPLEX; Biochemistry & Molecular Biology; Structure based drug design; Molecular docking; Technology; DRUG DISCOVERY; Computer Science; PROTEIN; Chemical descriptor; Chemistry; NNRTI; INDUCED FIT; Life Sciences & Biomedicine; Chemistry, Multidisciplinary; FEATURES; Biophysics
• Research Categories: Chemistry, Biochemistry; Biology, Molecular

Abstract

Non-nucleoside reverse transcriptase inhibitors (NNRTI) are allosteric inhibitors of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT), a viral polymerase essential to infection. Despite the availability of >150 NNRTI-bound RT crystal structures, rational design of new NNRTI remains challenging because of the variability of their induced fit, hydrophobic binding patterns. Docking NNRTI yields inconsistent results that vary markedly depending on the receptor structure used, as only 27% of the >20k cross-docking calculations we performed using known NNRTI were accurate. In order to determine if a hospitable receptor for docking could be selected a priori, we evaluated more than 40 chemical descriptors for their ability to pre-select a best receptor for NNRTI cross-docking. The receptor selection was based on similarity scores between the bound- and target-ligands generated by each descriptor. The top descriptors were able to double the probability of cross-docking accuracy over random receptor selection. Additionally, recall of known NNRTI from a large library of similar decoys was increased using the same approach. The results demonstrate the utility of pre-selecting receptors when docking into difficult targets. [Figure not available: see fulltext.]