Two of our knowledge-based scoring functions in MDock, ITScore and STScore, were evaluated on the 2011 CSAR Benchmark blind prediction exercise. ITScore was derived by using a statistical mechanics-based iterative method to circumvent the reference state problem. STScore was derived by accounting for the sparse data problem using a statistical approach. Our binding mode analysis showed that redocking the native ligands into their co-crystallized protein structures yielded success rates of 91.18%, 100%, and 100% with the criterion of rmsd < 2 A when the top 1, 2, and 3 poses were considered, respectively, suggesting the accuracy of the scoring functions and orientational sampling. We also analyzed the effects of a variety of factors on our mode predictions, suggesting that how well the ligand conformations are sampled based on the SMILES strings has the most critical effect. We performed comparative studies using UCSF DOCK. The conclusions would help improve docking algorithms.

Exercise on 2011 CSAR Benchmark using the iterative knowledge-based scoring functions ITScore and STScore of the MDock software: What we have learned

244th National Meeting (2012)

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