Due to the rising number of solved protein structures, computer-based techniques for automatic protein function annotation and classification into families are of high practical interest. In this work, we combine the automated pocket detection and classification tool DoGSite1 and the structure-based design platform LeadIT2 for protein assessment. Function predictions with DoGSite are based on global and local descriptors of automatically predicted pockets, incorporated into a support vector machine and a nearest neighbor search, respectively. In parallel, LeadIT is used to screen selected reference molecules against a protein. Function predictions are based on achieved HYDE4 scores of a specific reference molecule class.As proof of concept, four enzyme classes (lipases, cytochromes, thiamine-diphosphate dependent enzymes and medium-chain dehydrogenases)5 comprising 943 structures and 187 reference molecules are exemplarily chosen. Both methods show high prediction accuracies. Currently, the methodology is applied to all available enzymes in the PDB. References:A.  Volkamer et al., JCIM, 2010 Nov 22;  50(11)www.biosolveit.de/leadit I.  Reulecke et  al., Chem. Med. Chem., 2008 Jun;3(6)http://www.itb.uni-stuttgart.de/research/bioinformatics.html The project is part of the cluster Biokatalyse2021 and funded by the BMBF under grant 0315292A.

Combining automatic active site analysis and docking for structure-based protein function predictions

243rd National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.