Optimization of a pharmacophore-based correlation vector descriptor for similarity searching

The usefulness of a correlation-vector based, topological pharmacophore descriptor for similarity searching was examined in matters of three different scaling methods and a fuzzy binning scheme. "Retrospective screening" using twelve different datasets was performed to evaluate the influence of these parameters by means of the enrichment factor. Significant enrichment of active compounds was yielded in this study. Whereas the application of the fuzzy binning scheme did not substantially influence the enrichment factor, we observed significant higher enrichment for individual ligand families with the employment of a specific descriptor scaling method. These results indicate that optimization of individual critical parameters of a well-established descriptor may be crucial for improved performance in similarity searching and focused library design.