Virtual screening (VS) approaches have been extensively used in the drug discovery process for hit identification, lead optimization, and scaffold hopping. Ligand based screening approaches provide fast and inexpensive alternatives to high-throughput screening (HTS); however, careful consideration of the VS approach is required as performance varies significantly for different targets and queries. In the present work, the Database of Useful Decoys – Enhanced (DUD-E) was screened with several 3D ligand-based search approaches (ROCS, Phase, Shapeflex, Surflex-Sim) using either a single structure or multiple query structures as input. An average hit rate of 1% (HR-1%) was determined and compared to that of 2D-similarity search MACCS fingerprints. Based on this framework, a novel approach for screening large chemical databases using the fusion of multiple ligand-shape fingerprints is proposed.
Results: Relative to similarity searches with MACCS fingerprints and single query, Surflex-Sim and ROCS improved the HR-1% by 13% and 8%, respectively, during screening of the DUD-E. But due to a severe performance dip for low 2D similarity, it proves to only be reliable when there are high 2D similarities. In contrast, Shapeflex and ROCS peak performance occurred for hits with low 2D similarity. Using multiple structures as an ensemble of queries significantly improved the hit rates. Thus, HT-1% increased by 24% and 19% for Shapeflex and Surflex-Sim, respectively, when five queries were used. Shapeflex ultimately surpassed all other methods, identifying 67% and 72% of hits at 1% using 10 and 15 queries, respectively, while Surflex-Sim’s efficiency was optimal using 5 queries and identified 51% of hits.
Conclusion: The multiple ligand’s 4D-fingerprint method of Shapeflex is by far the most efficient screening approach benchmarked in this study. Enhanced hit enrichment of shape-based screening with multiple query structures, especially the fusion of multiple ligand fingerprints within Shapeflex, was revealed for the first time herein.

adel hamza

Appearances