A practical application of two in silico systems for identification of potentially mutagenic impurities.

摘要: The International Conference on Harmonization (ICH) M7 guidance for the assessment and control of DNA reactive impurities in pharmaceutical products includes use silico prediction systems as part hazard identification risk strategy. This is first internationally agreed document to include these types approaches. guideline requires two complementary approaches, an expert rule-based method a statistical algorithm. In addition, states that output from computer-based assessments can be reviewed using knowledge provide additional support or resolve conflicting predictions. approach designed maximize sensitivity correctly identifying compounds while providing framework reduce number need synthesized, purified subsequently tested Ames assay. Using data set 801 chemicals intermediates, we have examined relative predictive performances some popular commercial are common across industry. overall accuracy each was fairly comparable ranging 68% 73%; however, system (i.e. how many positive identified) varied much more dramatically 48% 68%. We explored combined under ICH enhance detection molecules. Finally, four smaller sets molecules, value review process, especially cases where disagreed their predictions, care when evaluating predictions large sets.