When the International Council for Harmonisation (ICH) published the ICH M7 guideline in 2014, it marked a major milestone for computational toxicology. For the first time, an internationally harmonized regulatory guideline recommended the use of computational (Q)SAR methods as part of a testing strategy for assessing mutagenic impurities in pharmaceuticals. Since then, updates to the guideline have continued to reinforce the important role of in silico prediction in supporting impurity risk assessments.¹
One of the key recommendations in ICH M7 is the use of two complementary (Q)SAR methodologies to predict the outcome of a bacterial mutagenicity assay. Rather than relying on a single model, the guideline recognizes that different computational approaches yield distinct types of evidence, thereby enabling a more robust assessment.
A computational toxicology assessment should be performed using (Q)SAR methodologies that predict the outcome of a bacterial mutagenicity assay… Two (Q)SAR prediction methodologies that complement each other should be applied. One methodology should be expert rule-based, and the second methodology should be statistical-based.¹
Learn more about how MultiCASE supports ICH M7 assessments with complementary expert rule-based and statistical (Q)SAR methodologies.
At first glance, using two computational models to answer the same question may seem redundant. In reality, each methodology approaches the problem differently, and each has its own strengths.
The combination of expert knowledge and statistical analysis helps improve confidence in prediction results while reducing the limitations associated with relying on a single modeling approach.
Expert rule-based models make predictions based on established structure-activity relationships (SARs). These relationships are built from decades of toxicological research and connect specific structural features within a molecule to known biological activity.
For ICH M7 assessments, expert rule-based models identify structural alerts associated with bacterial mutagenicity, including well-established alerts such as those published by Ashby and Tennant.² Rather than learning from data alone, these models apply expert knowledge to determine whether a compound contains features associated with mutagenic potential.
Within CASE Ultra, this approach is implemented through GT_EXPERT, the expert rule-based model used to support bacterial mutagenicity assessments under ICH M7.
Statistical models approach the same problem from a different perspective.
Instead of relying on predefined expert rules, statistical (Q)SAR models analyze large datasets of compounds with known biological activity to identify relationships between chemical structure and mutagenicity. Using computational methods such as machine learning and statistical analysis, these models recognize patterns that correlate with active and inactive compounds.³
Because they are data-driven, statistical models can identify both structural features associated with mutagenicity and features that may reduce mutagenic potential.⁴
Within CASE Ultra, GT1_BMUT serves as the statistical model used for ICH M7 bacterial mutagenicity predictions.
| Expert Rule-Based | Statistical (Q)SAR |
|---|---|
|
Based on expert knowledge |
Based on experimental data |
|
Uses structural alerts |
Learns from large datasets |
|
Transparent reasoning |
Pattern recognition |
The ICH M7 guideline recommends combining expert rule-based and statistical methodologies because they complement one another.⁵ Expert rule-based models contribute decades of scientific knowledge and transparent mechanistic reasoning, while statistical models contribute insights learned directly from experimental data.
Like laboratory assays, validated computational models may occasionally produce different results due to differences in methodology, training data, algorithms, and model design. Using two complementary approaches provides additional context and helps strengthen the overall assessment.⁶
Rather than asking which model is “better,” ICH M7 recognizes that both approaches contribute valuable evidence. Together, they provide a more comprehensive evaluation of mutagenic risk than either methodology could achieve alone.
(1) ICH. M7(R2) Assessment and Control of DNA Reactive (Mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk. https://database.ich.org/sites/default/files/ICH_M7%28R2%29_Guideline_Step4_2023_0216_0.pdf.
(2) Ashby, J.; Tennant, R. W. Definitive Relationships among Chemical Structure, Carcinogenicity and Mutagenicity for 301 Chemicals Tested by the U.S. NTP. Mutat. Res. Genet. Toxicol. 1991, 257 (3), 229–306. https://doi.org/10.1016/0165-1110(91)90003-E.
(3) Gini, G. QSAR Methods. In In Silico Methods for Predicting Drug Toxicity; Benfenati, E., Ed.; Springer US: New York, NY, 2022; pp 1–26. https://doi.org/10.1007/978-1-0716-1960-5_1.
(4) Jayasekara, P. S.; Skanchy, S. K.; Kim, M. T.; Kumaran, G.; Mugabe, B. E.; Woodard, L. E.; Yang, J.; Zych, A. J.; Kruhlak, N. L. Assessing the Impact of Expert Knowledge on ICH M7 (Q)SAR Predictions. Is Expert Review Still Needed? Regul. Toxicol. Pharmacol. 2021, 125, 105006. https://doi.org/10.1016/j.yrtph.2021.105006.
(5) Sutter, A.; Amberg, A.; Boyer, S.; Brigo, A.; Contrera, J. F.; Custer, L. L.; Dobo, K. L.; Gervais, V.; Glowienke, S.; Gompel, J. van; Greene, N.; Muster, W.; Nicolette, J.; Reddy, M. V.; Thybaud, V.; Vock, E.; White, A. T.; Müller, L. Use of in Silico Systems and Expert Knowledge for Structure-Based Assessment of Potentially Mutagenic Impurities. Regul. Toxicol. Pharmacol. 2013, 67 (1), 39–52. https://doi.org/10.1016/j.yrtph.2013.05.001.
(6) Kruhlak, N. L.; Benz, R. D.; Zhou, H.; Colatsky, T. J. (Q)SAR Modeling and Safety Assessment in Regulatory Review. Clin. Pharmacol. Ther. 2012, 91 (3), 529–534. https://doi.org/10.1038/clpt.2011.300.
