Similarity Thresholds

The similarity threshold represents the desired upper and lower bounds for the Tanimoto Coefficient (Tc). The Tc represents one measure of structural similarity between compounds. The Tc ranges from 1.0 (highest homology) to 0.0 (no homology). Extensive experience with this similarity measure suggests that a range from 1.0 to 0.70 (the default values in Neighbor_Search) seems optimal for a first pass. If too many compounds are found using these values, the lower bound should be raised from 0.70. Likewise, if too few compounds are found, the range may be expanded by decreasing the lower bound. Research at Abbott Laboratories with the Tc/Daylight fingerprint suggests that, on average, relatively high homology (Tc >= 0.85) is required in order to see similarity in terms of biological activity.