Identification of cavities on protein surface using multiple computational approaches for drug binding site prediction

• DOI: 10.1093/bioinformatics/btr331
• Publication Date: 2011-06-03T11:53:17Z
• Authors: Zengming Zhang; Yu Li; Biaoyang Lin; Michael Schroeder; Bingding Huang
• Keywords: DrugBank; Identification

Abstract Motivation: Protein–ligand binding sites are the active on protein surface that perform functions. Thus, identification of those is often first step to study functions and structure-based drug design. There many computational algorithms tools developed in recent decades, such as LIGSITEcs/c, PASS, Q-SiteFinder, SURFNET, so on. In our previous work, MetaPocket, we have proved it possible combine results methods together improve prediction result. Results: Here, continue work by adding four more Fpocket, GHECOM, ConCavity POCASA further success rate. The new method MetaPocket 2.0 individual approaches all tested two datasets 48 unbound/bound 210 bound structures used before. show average rate has been raised 5% at top 1 compared with work. Moreover, construct a non-redundant dataset drug–target complexes known structure from DrugBank, DrugPort PDB database apply this predict sites. As result, >74% target correctly identified 3 prediction, 12% better than best approach. Availability: web service test freely available http://projects.biotec.tu-dresden.de/metapocket/ http://sysbio.zju.edu.cn/metapocket. Contact: [e-mail] Supplementary Information: data Bioinformatics online.