Risk-based in silico mutagenic assessment of benzodiazepine impurities using three QSAR tools

Abstract

Benzodiazepines, widely prescribed psychoactive drugs, may contain DNA-reactive (mutagenic) impurities formed during synthesis, posing significant health risks. Owing to animal testing requirements, traditional in vitro and in vivo methods for assessing mutagenicity are time-consuming, costly, and ethically challenging. Computational approaches, particularly in silico (Q)SAR models, provide an efficient alternative for predicting toxicity based on chemical structure. This study evaluated the mutagenic potential of 88 benzodiazepine-related impurities using three freely accessible (Q)SAR tools: TOXTREE (Ames Test Alert by ISS), Toxicity Estimation Software Tool (TEST) with nearest neighbour and consensus models, and VEGA, a QSAR tool that integrates multiple mutagenicity prediction models, including the CAESAR Ames Mutagenicity Model. The tools were validated using a dataset of 99 chemicals with known Ames test results. TOXTREE exhibited the highest sensitivity (80.7 %) and accuracy (72.2 %) for predicting mutagenicity, whereas VEGA and TEST provided balanced accuracy (66.2 % and 66.7 %, respectively) and high specificity (74.5 % and 76.6 %, respectively). The risk assessment categorised 21 impurities as high risk, 11 as moderate-high risk, 28 as moderate-low risk, 22 as low risk, and 6 as equivocal, with expert review finalising classifications. The findings emphasise the integration of multiple (Q)SAR tools for early mutagenicity detection, regulatory compliance, and reduced reliance on animal testing. Further refinement of predictive models and additional computational approaches are recommended to enhance the accuracy of the risk assessment.

Keywords

• Benzodiazepines
• Mutagenic impurities
• In silico evaluation
• (Q)SAR tools
• Regulatory implications