Hao Tian: Prediction of allosteric sites through ensemble learning

Allostery is the process by which proteins transmit perturbations caused by the binding effect at one site to another distal site. Allostery is considered important in regulating protein's activity. Drug development depends on the understanding of allosteric mechanisms, especially the identification of allosteric sites, which is prerequisite in drug discovery and design. Many computational methods have been developed for allosteric site prediction using pocket features and dynamics information. Here, we provide a novel ensembled model, consisting of eXtreme gradient boosting (XGBoost) and graph convolutional neural network (GCNN) to predict allosteric sites. Our model can learn both physical properties and topology structure without any prior information and exhibited good performance under several indicators. Prediction results have shown that 84.9% of allosteric pockets in the testing proteins appeared in the top 3 positions. An interactive and quick-response server, PASSer: Protein Allosteric Sites Server (https://passer.smu.edu), is provided to facilitate further analysis in drug discovery.