Comparison of Fingerprints by Analyses of Bits Usage

Abstract

In the field of chemical information, various structural descriptors have been developed for representing chemical compounds. The fingerprint method is one of the structural descriptor by using a Boolean array representing the presence or absence of chemical substructures. As the processing speed for calculating the similarity score between two Boolean array is fast, the fingerprint method is used in similarity search of chemical compounds and in clustering of chemical structures for the screening. There are various definitions in fingerprint method, because characteristics are represented by the difference of chemical substructure. It is important to select the definition of the fingerprint method in consideration of data sets and in obtaining expected results. To evaluate the features of those definitions, we evaluated the use of bits by analysis using the several values in the database, and evaluated the comparison of retrieval results.