Subsimilarity searching

The subsimilarity search method developed at the Upjohn Company, which is discussed later, measures the similarity between a pair of molecules by the maximal common substructure (MCS), that is, the largest substructure that is common to the two molecules that are being compared, as illustrated in Figure 4. The system utilizes an inverse-frequency weighting scheme that favors heteroatom-heteroatom pairs over the mote frequently occurring carbon-carbon pairs. For each pair of heteroatoms separated by minimum paths of two to six bonds, an additional bond is added to link the atoms directly. This additional link results in an increase in the size of the MCS and thus a greater similarity for heteroatom-containing common substructures. 

The linking of fragment-based and MCS-based searching provides an extremely powerful mechanism for subsimilarity searching (and could clearly be extended to both similarity and supersimilarity searching) with developments in computing power, it is likely that such systems will become widely used over the next few years. 

By definition, any target structure that is a substructure (or superstructure) of the database molecule will have a subsimilarity (or supersimilarity) of 1.0, and hence a subsimilarity (supersimilarity) search can be viewed as a form of fuzzy substructure (superstructure) search. A further difference from the systems described thus far is in the nature of the output. The description of similarity searching that has been given previously assumes that a user browses down a ranked output until sufficient database structures have been identified. Here, conversely, the user specifies a threshold similarity value, and the output consists of an unordered list of all database structures (or database reactions) that have a similarity to the target that exceeds the chosen threshold value. 

It will be clear from the material presented earlier that 2D similarity searching (either in its basic form or in its subsimilarity or supersimilarity forms) has become a standard retrieval option in chemical information systems... 

---