• Elena Fioravanzo

ICH-M7 QSAR/Expert Judgment Workshop - Tokyo

Chihae Yang, CEO of MN-AM, will give a talk on "Combining evidence within and beyond the confines of (Q)SAR to predict mutagenicity" at the M7/QSAR Workshop organized in Tokyo in November 2019 by Masayuki Mishima (Chugai Pharmaceutical). Prof. Mark Cronin is a co-author.

Do you want to receive a copy of the presentation or to know more? Send a request to elena@toxnavigation.com.


Robust in silico toxicity prediction enables identification of chemicals with potential adverse effects while reducing the need for experimental studies. Accurately estimating the likelihood of a chemical being genotoxic continues to be important across all industrial sectors and is crucial in regulatory programs including ICH M7. Quantitative structure-activity relationships (QSARs) and expert knowledge-based rules are commonly employed for in silico models and are well-suited for DNA-reactive mutagenicity where modes of chemical reactivity are well characterized. While much effort is being dedicated to the in silico modelling of one particular endpoint, bacterial reverse mutagenesis, there still remains a need to establish a systematic method to consider and incorporate other relevant pieces of evidence including, for example, experimental results from mechanistic analogs to improve confidence in predictions. This talk will present new strategies for combining one or more (Q)SAR predictions, with an option to include experimental analog data when appropriate. These methods quantitatively take into account the reliability of each evidence source. The reliability of a (Q)SAR prediction can be estimated from model validation, statistics, and applicability domain analysis. The reliability of experimental data is derived from expert evaluation of study protocols and results, and also the suitability of each analog from structural, physicochemical, and biological perspectives. All or some of the selected pieces of evidences are combined within probability bounds or using Bayesian network approaches. Reflecting on the 60-year trajectory of QSAR approaches, this talk will emphasize how toxicity prediction in the future will be improved by moving beyond the confines of (Q)SAR by making it a more integrated predictive science.