The three dimensional structure of proteins is the key to understanding their function and evolution. For a substantial proportion of proteins in the Protein Data Bank, their functions are not known. However, given the coordinates of protein atoms in the Protein Data Bank, we can construct models of protein graphs and apply clique algorithms to search for maximal common subgraphs (MCS) to detect tertiary structural resemblance in proteins. Since binding sites are evolutionarily conserved across very distant protein families, finding similar binding sites between known and unknown proteins can provide clues as to the functions of unknown proteins.

We introduce a novel strategy of binding site comparison, unknown vs known protein functions, to construct a hypothetical protein-ligand complex. Subsequently, molecular dynamics has validated that this protein most likely binds and repairs the damaged DNA similar to known DNA-repair enzymes. Our methodology is general and enables one to determine functions of other proteins currently labelled as “unknown function”. The methodology we propose, the binding sites comparisons, enhanced by molecular dynamics, looks promising. It will not only stimulate the function prediction of other uncharacterized proteins with structures in the Protein Data Bank, it will also boost experimental studies of proteins whose function is still not known.

Janez Konc is an Assistant Professor of Medicinal Chemistry at University of Ljubljana, Slovenia and a Senior Fulbright Scholar at the NIH. He received a PhD in Pharmacy from the University of Ljubljana in 2008. His current research interests include development of algorithms for NP hard problems (http://www.sicmm.org/~konc/maxclique/),  development of bioinformatics algorithms using protein graph theory,  development of web servers,  e.g.,  ProBiS (http://probis.cmm.ki.si) for the detection of protein binding sites and ENZO (http://enzo.cmm.ki.si) for the study of enzyme kinetics.