The wealth of protein and complex structural information that is searchable and readily available has greatly facilitated the drug discovery process. Within such data assessment of binding site similarities have been used for a number of purposes including to assess the risk of off-target effects for a potential ligand, design of ligands that are selective for targets of interest, and in knowledge-based bioisostere identification. In this context ligand similarity or protein residue features are used to characterize and identify similarity, however we believe there are advantages to focusing on characteristics of the binding site cavity. Given that a significant fraction of the structures in the PDB represent APO macromolecular structures, such cavity-based approaches are expected to find hits that would otherwise be missed.. We describe a new approach to assess binding site similarity facility that extends upon ideas in the IsoMIF approach. Interaction fields are computed indicating the most probable location of features such as hydrophobicity, aromaticity, hydrogen bonding, and charged groups in the binding site cavity. Representation of the binding site as a collection of features enables very rapid calculation of binding site pairwise similarities and alignments using our GPU-enabled Shape Screening technology that is &gt;100x faster than computation on the CPU. Key findings are reported from extensive validation of the methodology against a number of widely used binding site similarity datasets. The potential of this new cavity-centric binding site similarity approach to off-target effect prediction, target identification, and knowledge-based bioisostere identification is shown.

GPGPU enhanced binding site similarity determination

256th National Meeting (2018)

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