Are Topological Properties of Drug Targets Based on Protein-Protein Interaction Network Ready to Predict Potential Drug Targets?

In this study, three levels of verification strategies were designed and conducted by applying support vector machine (SVM) as the classification tool. Unfortunately, none of the models built was capable to get the accuracy better than 80%, and most of the performances were between 0.6-0.7 for the area under the curve (AUC). Hence, conclusion was proposed that topological properties of drug targets are not prepared to predict potential drug targets based on the present PPIN. This summary was also underpinned by the unsatisfying prediction data reported by other investigators. Finally, data quality and quantity of the protein-protein interactions (PPI) are being expected to be improved and ameliorated.
PMID: 26552443 [PubMed - as supplied by publisher]

Source: Combinatorial Chemistry and High Throughput Screening - November 10, 2015 Category: Chemistry Authors: Li S, Yu X, Zou C, Gong J, Liu X, Li H Tags: Comb Chem High Throughput Screen Source Type: research