A new bioinformatic approach to detect common 3D sites in protein structures

• DOI: 10.1002/prot.10339
• Publication Date: 2003-06-24T10:16:06Z
• Authors: Martin Jambon; Anne Imberty; Gilbert Deléage; Christophe Geourjon
• Keywords: Protein superfamily

Abstract An innovative bioinformatic method has been designed and implemented to detect similar three‐dimensional (3D) sites in proteins. This approach allows the comparison of protein structures or substructures detects local spatial similarities: this is completely independent from amino acid sequence backbone structure. In contrast already existing tools, basis for a representation structure by set stereochemical groups that are defined independently notion acid. efficient heuristic finding similarities uses graphs triangles chemical represent developed. The implementation constitutes software named SuMo (Surfing Molecules), which dynamic definition groups, selection proteins, management screening databases. To show relevance approach, we focused on two extreme examples illustrating convergent divergent evolution. unrelated serine proteases, one common site, corresponds catalytic triad. legume lectins family composed >100 share sequences folds but may have lost their ability bind carbohydrate molecule, discriminates between functional non‐functional with selectivity 96%. time needed searching given site typically 0.1 s PIII 800MHz/Linux computer; thus, further studies, will be used screen PDB. Proteins 2003;52:137–145. © 2003 Wiley‐Liss, Inc.