Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6451301 | Computational Biology and Chemistry | 2017 | 6 Pages |
â¢A new R-subspace alternative to protein-cavity is proposed to localize binding sites.â¢R-subspace is found to be better to localize ligand-binding site (LBS).â¢Proteins for which cavity-subspace fails to localize LBS can be predicted.â¢R-subspace compensates most of the cavity-failure cases.â¢R-subspace and cavity-subspace complementarily enhance success rate to localize LBS.
The common exercise adopted in almost all the ligand-binding sites (LBS) predictive methods is to considerably reduce the search space up to a meager fraction of the whole protein. In this exercise it is assumed that the LBS are mostly localized within a search subspace, cavities, which topologically appear to be valleys within a protein surface. Therefore, extraction of cavities is considered as a most important preprocessing step for finally predicting LBS. However, prediction of LBS based on cavity search subspace is found to fail for some proteins. To solve this problem a new search subspace was introduced which was found successful to localize LBS in most of the proteins used in this work for which cavity-based method MetaPocket 2.0 failed. Therefore this work appeared to augment well the existing binding site predictive methods through its applicability for complementary set of proteins for which cavity-based methods might fail. Also, to decide on the proteins for which instead of cavity-subspace the new subspace should be explored, a decision framework based on simple heuristic is made which uses geometric parameters of cavities extracted through MetaPocket 2.0. It is found that option for selecting the new or cavity-search subspace can be predicted correctly for nearly 87.5% of test proteins.
Graphical abstractDownload high-res image (211KB)Download full-size image