|
[摘要]:A novel scoring algorithm based on mol. interaction fingerprints (IFPs) was comparatively evaluated in its scaffold hopping efficiency against four virtual screening stds. (GlideXP, Gold, ROCS, and a Bayesian classifier). Decoy databases for the two targets under examn., adenosine deaminase and retinoid X receptor alpha, were obtained from the Directory of Useful Decoys and were further enriched with approx. 5% of active ligands. Structure and ligand-based methods were used to generate the ligand poses, and a Tanimoto metric was chosen for the calcn. of the similarity interaction fingerprint between the ref. ligand and the screening database. Database enrichments were found to strongly depend on the pose generator algorithm. In spite of these dependencies, enrichments using mol. IFPs were comparable to those obtained with GlideXP, Gold, ROCS, and the Bayesian classifier. More interestingly, the mol. IFP scoring algorithm outperformed these methods at scaffold hopping enrichment, regardless of the pose generator algorithm. |
|