You are here:
Toxico-Cheminformatics and QSPR Modeling of the Carcinogenic Potency Database
Citation:
Wang, K., A. Tropsha, A. M. RICHARD, AND I. Rusyn. Toxico-Cheminformatics and QSPR Modeling of the Carcinogenic Potency Database. Presented at International Science Forum on Computational Toxicology, Research Triangle Park, NC, May 21 - 23, 2007.
Impact/Purpose:
Report on the development of a tiered, confirmatory scheme for prediction of chemical carcinogenicity based on QSAR studies of compounds with available mutagenic and carcinogenic data.
Description:
Report on the development of a tiered, confirmatory scheme for prediction of chemical carcinogenicity based on QSAR studies of compounds with available mutagenic and carcinogenic data. For 693 such compounds from the Carcinogenic Potency Database characterized molecular topological descriptors, k-nearest-neighbor QSAR models were developed for (i) mutagens vs. non-mutagens, (ii) carcinogens vs. non-carcinogens, (iii) genotoxic carcinogens vs. non-genotoxic carcinogens, and (iv) genotoxic carcinogens vs. genotoxic non-carcinogens. Except for case (ii), models with predictive accuracy exceeding 80% each were obtained for training, test and external validation sets. Our results compare favorably with those generated with more common fragment-based approaches. Our method can reduce the rate of false positives/negative and strengthen the prediction performance of the QSAR models. This concept can be generalized and extended to better integrate other types of carcinogenicity characteristics in aiding classification, e.g., tumor sites, TD50 range, multisite, multisex, multispecies tumor incidences, etc.