A retrospective analysis of pharmaceutical regulatory data found better performance for Ames-test (Q)SAR models used in ICH M7 work than for models of mammalian-cell or in vivo genotoxicity. Reliability checks improved performance in some comparisons.
Researchers affiliated with medicines agencies in Austria and France and with Austrian universities assembled experimental results from public and proprietary pharmaceutical regulatory procedures. They used the curated data to examine available (quantitative) structure–activity relationship, or (Q)SAR, models for bacterial mutagenicity in the Ames test, genotoxicity in mammalian cells in vitro, and genotoxicity in vivo.
The team compared predictions across those experimental endpoints and examined structural alerts, chemical subspaces, prediction-reliability filters and expert review as ways to improve model performance.
The authors found that performance depended on both the endpoint and the model. Ames-test models routinely used in ICH M7 analyses performed better than the models assessed for mammalian-cell and in vivo genotoxicity.
The additional analyses improved prediction performance in many comparisons. Even so, the authors concluded that mammalian-cell and in vivo genotoxicity models require further refinement before they can be used in regulatory decision-making.
A medRxiv preprint simulated rosuvastatin exposure during ticagrelor coadministration in four age–ABCG2 genotype groups. Exposure-matching results differed by group, but the proposed dose scenarios were not clinically tested.
NIH researchers compared ADME@NCATS predictions with laboratory measurements for selected Tox21 compounds. Three basic ADME models reached ROC AUCs of 0.81–0.92; a separate passive blood–brain barrier model reached 0.77. The work supports compound triage, not validated prediction of in vivo toxicity.
