3DSTarPred: A Web\nServer for Target Prediction of\nBioactive Small Molecules Based on 3D Shape Similarity

Target identification plays a critical\nrole in preclinical\ndrug\ndevelopment. The <i>in silico</i> approach has been developed\nand\nwidely applied to assist medicinal chemists and pharmacologists in\ndrug target identification. There are many target prediction web servers\navailable today that have revealed both advantages and shortcomings\nin practical applications. Here, we present 3DSTarPred, a web server\nfor three-dimensional (3D) shape similarity-based target prediction\nof small molecules. A benchmark study showed that 3DSTarPred achieved\na target prediction success rate of 76.27%, which was higher than\nthat of existing target prediction web servers. In addition, the performance\nof 3DSTarPred in the target prediction of diverse substructures/superstructures\nwas also better than that of the existing target prediction web servers.\nIn case studies, 3DSTarPred was used to identify the potential targets\nof two small molecules, one being kaempferol, a natural lead compound\nfor the treatment of Alzheimer’s disease (AD), and the other\nbeing sildenafil, a candidate for drug repurposing in AD. The case\nstudies further demonstrated the reliability and success of 3DSTarPred\nin practice. The 3DSTarPred server is freely available at http://3dstarpred.pumc.wecomput.com.