||Computer-Assisted Structure Activity Relationships of Nitrogenous Cyclic Compounds Tested in Salmonella Assays for Mutagenicity.
Walsh, D. B. ;
Claxton, L. D. ;
||Health Effects Research Lab., Research Triangle Park, NC. ;Massachusetts Rehabilitation Commission, Boston.
Nitrogen heterocyclic compounds ;
Computer applications ;
Molecular structure ;
Salmonella typhimurium ;
Pattern recognition devices ;
||Most EPA libraries have a fiche copy filed under the call number shown. Check with individual libraries about paper copy.
Study of the relationship between mutagenicity and molecular structure for a data set of nitrogenous cyclic compounds is reported. A computerized SAR system (ADAPT) was utilized to classify a data set of 114 nitrogenous cyclic compounds with 19 molecular descriptors. All of the descriptors represented at least 10% of the compounds in the data sets. The average correct predictability of the data base was calculated to be 89% after evaluating 100 training/prediction subsets. The actual predictive ability of the discriminants generated by the ADAPT system was demonstrated by predicting the mutagenicity of structurally similar compounds not in the data set. Weight vectors generated in the pattern recognition programs were used to predict the bacterial mutagenicity of 10 compounds which were not included in the data set. All of the compounds were predicted correctly which was actually better than the 89% calculated by the system. The authors displayed the ability of the system of classify compounds as mutagenic or nonmutagenic by using molecular descriptors and to predict the biological activity of untested chemicals with a high degree of confidence. The paper presents the uses of the type of SAR approach in a research laboratory. (Copyright (c) 1987 Elsevier Science Publishers (Biomedical Division).)