Molecular docking of small molecules in the protein binding sites is the
most widely used computational technique in modern structure-based drug discovery. Although
accurate prediction of binding modes of small molecules can be achieved in most cases, estimation
of their binding affinities remains mediocre at best. As an attempt to improve the correlation
between the inhibitory constants, pKi, and scoring, we created a new, hybrid scoring function. The
new function is a linear combination of the terms of the scoring functions of AutoDock and
AutoDock Vina. It was trained on 2,412 protein-ligand complexes from the PDBbind database
(www.pdbbind.org.cn, version 2012) and validated on a set of 313 complexes released in the 2013
version as a test set. The new function was included in a modified version of AutoDock. The hybrid
scoring function showed a statistically significant improvement in both training and test sets in
terms of correlation with and root mean square and mean absolute errors in prediction of pKi values.
It was also tested on the CSAR 2014 Benchmark Exercise dataset (team T) and produced
reasonably good results.
Keywords: AutoDock, AutoDock Vina, hybrid scoring function, molecular docking, scoring, virtual
Rights & PermissionsPrintExport