Identifying calcium-binding sites with oxygen-carbon shell geometric and chemic criteria - A graph-based approach


Identifying calcium-binding sites in proteins help acknowledge protein functions. We thus developed a graph theory and geometry approach to improve the accuracy for predicting calcium-binding sites based on the valence function. In order to explore in-depth the geometric properties of the oxygen-carbon shells of calcium-binding sites, we enhance our previous approach at a high level to find biggest local oxygen clusters with a graph algorithm to find maximal cliques and propose a new geometric criterion embedding the bidentate property to filter non calcium-binding oxygen clusters. In addition, we apply some motifs of residue combinations as another filter to exclude non calcium-binding oxygen clusters possibly formed by hydrogen bonds and obtain higher site selectivity without trading off site sensitivity. The experiments demonstrate good predictive performance on both old and new datasets.


Computer Science

Document Type

Conference Proceeding



Publication Date


Journal Title

Proceedings - IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2008