Similarity searching in the Chemical Abstracts Service Registry File

摘要

Chemical Abstracts Service (CAS) has begun exploring chemical similarity or 'fuzzy' search techniques to complement the more traditional two-dimensional (2D) substructure search. Experimental software is being developed to rank a large set of compounds such as the Registry File topologically, relative to a target structure. Those with the highest ranking are assumed more closely to resemble an input target structure. This topological similarity ranking is based upon the number of substructural fragments in common between the target and file structures. Large answer sets can be further refined by using a secondary generic substructure search. Results of such searches on several thousand structure test subsets will be presented. In addition, analysis of the search results using the CA File as a retrospective 'experimental' source will be discussed.