Structure-activity relationship analysis tools: validation and applicability in predicting carcinogens

Area of Science:

• Toxicology
• Computational Chemistry
• Risk Assessment

Background:

• Carcinogenicity prediction is crucial for public health and regulatory decisions.
• Structure-activity relationship (SAR) software and genetic toxicity tests are common methods for assessing chemical carcinogenicity.
• Existing methods require validation and comparison for optimal use in risk assessment.

Purpose of the Study:

• To validate and compare the predictivity of OncoLogic, MultiCASE (MCASE), Ashby-Tennant structural alerts, and genetic toxicity testing for carcinogenicity.
• To evaluate the sensitivity and specificity of these methods, individually and in combination.
• To assess the performance of these methods for potent carcinogens.

Main Methods:

• Utilized 650 compounds from the Carcinogenic Potency Database (CPDB) with known carcinogenicity results.
• Compared the predictive accuracy of OncoLogic (v4.1), MCASE (v3.1), Ashby-Tennant alerts, and genetic toxicity data.
• Examined sensitivity for carcinogen prediction and specificity for non-carcinogen prediction.

Main Results:

• SAR analysis programs and structural alerts demonstrated good performance, particularly for compounds with low human exposure.
• These computational methods showed potential to supplement traditional genetic toxicity testing.
• Each method exhibited limitations in its applicability and predictive scope.

Conclusions:

• SAR analysis tools and structural alerts are valuable for predicting carcinogenicity, especially in a weight-of-evidence framework.
• These computational approaches can enhance the assessment of chemical carcinogenicity alongside genetic toxicity data.
• Careful consideration of method-specific limitations is necessary for accurate carcinogenicity prediction.