Prediction of protein-ligand binding affinities plays an essential role for molecular recognition and virtual screening. We have developed a scoring function, namely GemAffinity, to predict binding affinities by using a stepwise regression method and 88 descriptors from 891 complex structures. GemAffinity consists of five descriptors, including van der Waals contacts; metal-ligand interactions; water effects; ligand deformation penalty; and conserved hydrogen-bonded residues. Experimental results indicate that GemAffinity is the best among 13 methods on a test set and can enrich screening accuracies on four sets. We believe that GemAffinity is useful for virtual screening and drug discovery.
|頁（從 - 到）||27-41|
|期刊||International Journal of Data Mining and Bioinformatics|
|出版狀態||已發佈 - 二月 2012|
ASJC Scopus subject areas
- Library and Information Sciences
- Information Systems
- Biochemistry, Genetics and Molecular Biology(all)