Precise computational prediction of interaction sites in 3-D protein structures

BIOL 232

Ying Wei, wei.y@neu.edu and Mary Jo Ondrechen, M.Ondrechen@neu.edu. Department of Chemistry and Chemical Biology, Northeastern University, Boston, MA 02115
Many methods for the prediction of interaction sites in protein 3D structures are not very selective and have low precision. This problem is addressed using THEMATICS, a simple computational method for identification of active sites in protein structures. Using a test set of 169 enzymes from the Catalytic Site Atlas, it is shown that THEMATICS can deliver precise, localized site predictions. Adjustment of the cut-off criteria can improve recall rates for catalytic residues with only a small sacrifice in precision. With a preferred cut-off value of 0.99, THEMATICS achieves a high success rate of 93% for interaction site prediction. Recall rates for catalytic residue prediction are similar to those of other structure-based methods, but with two- to five- fold better precision and a low false positive rate of only 3%. This good performance is observed across the six E.C. classes and for unbound as well as bound structures.
 

Frontiers in Chemical Biology
5:00 PM-7:00 PM, Wednesday, August 22, 2007 BCEC -- Exhibit Hall - B2, Poster

Division of Biological Chemistry

The 234th ACS National Meeting, Boston, MA, August 19-23, 2007