Which carcinogenicity study should I use? Automated identification of reliable studies

Area of Science:

• * Pharmaceutical toxicology
• * Regulatory science
• * Computational chemistry

Background:

• * Deriving acceptable intakes for Class 1 mutagenic impurities (ICH M7) relies on selecting robust carcinogenicity studies.
• * Current study selection methods are often subjective, inconsistent, and time-consuming.
• * This process is critical for regulatory submissions and ensuring drug safety.

Purpose of the Study:

• * To develop and validate an automated workflow for identifying the most reliable carcinogenicity study.
• * To enhance the Lhasa reliability grading scheme with an A+ category for dose-response modeling suitability.
• * To support reproducible and non-animal-based decision-making frameworks in drug development.

Main Methods:

• * Implementation of an automated workflow within the LCDB Plus platform.
• * Extension of the Lhasa reliability grading scheme to include an A+ category.
• * Application of a Pareto optimisation algorithm to identify the most robust study.
• * Integration of the Lhasa TD50 model with Akaike Information Criterion and Pearson's χ2 goodness-of-fit test.

Main Results:

• * The automated workflow successfully identified a single most robust study in 97.6% of cases across 18,167 studies.
• * Selections aligned with expert-derived choices and published regulatory acceptable intakes.
• * The A+ category effectively distinguished studies suitable for dose-response modeling.

Conclusions:

• * The automated workflow significantly reduces the burden of carcinogenicity study selection.
• * It promotes reproducible, non-animal decision-making aligned with the 3Rs principles.
• * The system supports consistent and reliable acceptable intake derivation for ICH M7 Class 1 impurities.