Premium
A computerized structure–activity correlation program for relating bacterial mutagenesis activity to chemical structure
Author(s) -
Tinker John F.
Publication year - 1981
Publication title -
journal of computational chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.907
H-Index - 188
eISSN - 1096-987X
pISSN - 0192-8651
DOI - 10.1002/jcc.540020304
Subject(s) - consistency (knowledge bases) , substructure , computer science , correlation , mathematics , data mining , computational biology , artificial intelligence , biology , engineering , geometry , structural engineering
Abstract The program employs a substructure recognition process based on the procedure of Hodes et al. Structural information is accepted from connection tables ( Chemical Abstracts Registry II or III) or MCC ‐ TSS ciphers. Typically in this program the basic substructure unit (ganglion) is an atom triplet. Data for correlation are provided by over 1000 results from mutagenesis tests run in the Ames Salmonella typhimurium test, obtained from the literature. Activity may be calibrated in two to eight levels. Statistical weights for ganglia are calculated from the program of Hodes et al. The results of the program reveal the dependence of biological data on chemical structure, the self‐consistency of the data, and the level of activity to be expected of an unknown structure, with an estimate of confidence. The value of the program was shown with 34 unknowns simultaneously tested for activity in the Ames test and submitted to structure–activity analysis. Of these 74% were correctly predicted for the level of activity, and 88% to within one level of activity. It is expected that experience with the program in a trial application setting will enable us to restrict the number of activity calibrations to two. This will allow a simple selection of an inactive or active prediction to be made.