A new highly specific method for predicting the carcinogenic potential of pharmaceuticals in rodents using enhanced...

This report describes in detail a new quantitative structure-activity relational expert system (QSAR-ES) method for predicting the carcinogenic potential of pharmaceuticals and other organic chemicals in rodents, and a beta-test evaluation of its performance. The method employs an optimized, computer-automated structure evaluation (MCASE) software program and new database modules which were developed under a Cooperative Research and Development Agreement (CRADA) between FDA and Multicase, Inc. The beta-test utilized 126 compounds with carcinogenicity studies not included in control database modules and three sets of modules, including: A07-9 (Multicase, Inc.), AF1-4 (FDA-OTR/Multicase, Inc.), and AF5-8 (FDA-OTR/proprietary). The investigation demonstrated that the standard MCASE(A07-9) system which had a small data-set (n = 319), detected few structure alerts (SA) for carcinogenicity (n = 17), and had poor coverage for beta-test compounds (51%). Conversely, the new, optimized FDA-OTR/MCASE(AF5-8) system had a large data-set (n = 934), detected many SA (n = 58) and had good coverage (94%). In addition, the study showed the standard MCASE(A07-9) software had poor predictive value for carcinogens and specificity for noncarcinogens (50 and 42%), detected many false positives (58%), and exhibited poor concordance (46%). Conversely, the new, FDA-OTR/MCASE(AF5-8) system demonstrated excellent predictive value for carcinogens and specificity for non-carcinogens (97%, 98%), detected only one false positive (2%), and exhibited good concordance (75%). The dramatic improvements in the performance of the MCASE were due to numerous modifications, including: (a) enhancement of the size of the control database modules, (b) optimization of MCASE SAR assay evaluation criteria, (c) incorporation of a carcinogenic potency scale for control compound activity and MCASE biophores, (d) construction of individual rodent gender- and species-specific modules, and (e) defining assay acceptance criteria for query and control database compounds.