The success of structure-based drug discovery methods, including virtual screening, depends on accuracy in the scoring functions used to rank-order and select compounds for experimental testing. We have previously implemented and validated a post-docking rescoring function called the footprint similarity score into the program DOCK6. The function ranks candidate ligands on how closely their per-residue protein binding interactions mimic those of a reference molecule with known activity. To date, it has been used to identify lead compounds that bind to drug targets HIV gp41 (Holden, et al. Boorg Med Chem Lett, 2012 , 22, 3011-3016) and fatty acid binding protein (Berger, et al. PloS One, 2009 , 7, e50968). Here, we present a new formulation of the footprint similarity score that distributes key protein residues over multiple grids, enabling the function to be used in on-the-fly virtual screening. In this work, we demonstrate that the function yields improved success compared to the standard DOCK energy function in pose identification experiments over a large test set of crystal ligand-protein complexes, and in enrichment experiments to three systems from the Directory of Useful Decoys database. We also present the application of the scoring function to the de novo construction of molecules for compatibility with HIV gp41. The function is available for use in the current release of DOCK6.

Implementation and evaluation of a multi-grid scoring function for structure-based design

247th National Meeting (2014)

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