Prediction Of The Protein-Ligand Binding Sites Based On Geometric Algorithm And Evolutionary Sequence Conservation

With the advent of post-genomic era, the protein expression rule and its biology function has become the major task that life science researchers are facing with. The interaction between protein and ligand as well as the relationship between protein structure and function are the core content and moreover, the prediction of protein binding sites can provide foundation in these researches. Meanwhile, the prediction of protein ligand-binding sites is crucial in structure-based drug design (SBDD), significant in computer-aided drug design (CADD) and the hot spots in bioinformatics.As the ligand-binding site is often located in the pocket-shape area on the protein surface, a number of geometric methods have been developed to predict the ligand-binding sites by detecting the pockets on the protein surface. Here we present a novel method named ConHull, in which we combine incremental convex hull and solvent accessible surface to identify atoms located in the pocket area of the protein surface in the first step, then use K-means Clustering to divide them into 7 Clusters as site candidates. Next, the identified pockets are ranked by their geometric volumes and finally the evolutionary sequence conservation of surrounding residues is taken into account to re-rank the first four largest pockets.In order to validate our method, we compare ConHull with the other three tools:LIGSITEcs, PASS, and SURFNET, and evaluate them on a well-known dataset of 210 protein-ligand complexes. Our method achieves a success rate of over 90% at the top 3 and performs slightly better than the other approaches in overall prediction. In conclusion, the prediction of protein binding sites is not merely a geometric problem. The pocket size is mainly a vital factor in this prediction, and moreover, the conservation score can lead to significant improvements of the binding site prediction.