Developing New Algorithms For Protein-ligand Binding Sites Prediction

Protein-ligand binding sites are active sites on protein surface that perform protein functions. Thus the first step of protein functional study or structure-based drug design is to identify such binding sites on protein surface and require the information of residues around binding sites. Many computational algorithms and tools have been developed in the recent decades to predict protein-ligand binding sites, such as LIGSITECSC, PASS, Q-SiteFinder and SURFNET. These algorithms first define protein-ligand binding sites on protein surface based on certain characters and then identify these binding sites according to the definitions. However, as we tested, these algorithms which are based on one or few characters of binding sites don’t perform very well. Here we developed a new generation of algorithm to prediction protein-ligand binding sites which combines these traditional algorithms together to make the definition of binding sites more precise to improve the prediction success rate. Here we name our prediction method as MetaPocket.In the first version of MetaPocket (MetaPocket1.0), there were four traditional algorithms of LIGSITECSC, PASS, Q-SiteFinder and SURFNET whose results were combined together to improve the success rate. In this work, we continued on MetaPocket and developed better ways of combining different single algorithm. We integrated four more traditional algorithms of Fpocket, ConCavity, GHECOM and POCASA into MetaPocket2.0, which improved the success rate12%than all the single methods at the top1prediction. To test the performance of MetaPocekt2.0on the prediction of druggability, we built a drug-target dataset which includes all approved drugs and their target proteins with known structures. This dataset could be a benchmark data-set for protein druggability prediction.We have built a powerful web server for MetaPocket2.0for other academic researchers. Moreover, we proposed a common design pattern in our meta web-server which is extremely fast, strongly fault-tolerant and easy to extend. MetaPocket2.0web server is free available to academic users at:http://sysbio.zju.edu.cn/metapocket.