Assessment of a Novel Scoring Method Based on Solvent Accessible Surface Area Descriptors

A novel scoring algorithm based on unique solvent accessible surface area (SASA) descriptors was comparatively evaluated for its database enrichment potential against the virtual screening (VS) methods GOLD and Glide. Several protein test cases, including adenosine deaminase and estrogen receptor alpha, were used for the evaluation. The structure-based VS method GOLD was used to generate the protein-ligand docking poses. These docking poses were then postprocessed with a protein-ligand interaction fingerprint metric. Next, the SASA descriptors were computed for each ligand and its respective protein in their bound/unbound states; a Bayesian model was learned with SASA descriptors and subsequently used to score the remaining ligands in the screening databases. Early database enrichments using SASA descriptors were found comparable or superior to those of GOLD and Glide. Moreover, SASA descriptors display an outstanding robustness to produce satisfactory early enrichments for a large variety of target classes. Based on these encouraging results, these novel topological descriptors constitute a valuable in silico tool in hit finding practices.