Pred-binding: large-scale protein-ligand binding affinity prediction

• Published in: Journal of enzyme inhibition and medicinal chemistry Vol. 31; no. 6; pp. 1443 - 1450
• Main Authors: Shar, Piar Ali, Tao, Weiyang, Gao, Shuo, Huang, Chao, Li, Bohui, Zhang, Wenjuan, Shahen, Mohamed, Zheng, Chunli, Bai, Yaofei, Wang, Yonghua
• Format: Journal Article
• Language: English
• Published: England Taylor & Francis 01.11.2016
• Subjects: Binding affinity prediction; drug target interaction; Ligands; Prednisolone - analogs & derivatives; Prednisolone - metabolism; Protein Binding; random forest; Support Vector Machine; Binding affinity prediction; drug target interaction; support vector machine; random forest
• ISSN: 1475-6366; 1475-6374
• DOI: 10.3109/14756366.2016.1144594

Drug target interactions (DTIs) are crucial in pharmacology and drug discovery. Presently, experimental determination of compound-protein interactions remains challenging because of funding investment and difficulties of purifying proteins. In this study, we proposed two in silico models based on support vector machine (SVM) and random forest (RF), using 1589 molecular descriptors and 1080 protein descriptors in 9948 ligand-protein pairs to predict DTIs that were quantified by K i values. The cross-validation coefficient of determination of 0.6079 for SVM and 0.6267 for RF were obtained, respectively. In addition, the two-dimensional (2D) autocorrelation, topological charge indices and three-dimensional (3D)-MoRSE descriptors of compounds, the autocorrelation descriptors and the amphiphilic pseudo-amino acid composition of protein are found most important for K i predictions. These models provide a new opportunity for the prediction of ligand-receptor interactions that will facilitate the target discovery and toxicity evaluation in drug development.