Structure-activity relationship models for rat carcinogenesis and assessing the role mutagens play in model...

Area of Science:

• Toxicology
• Computational Chemistry
• Drug Discovery

Background:

• Structure-activity relationship (SAR) models are crucial for predicting chemical carcinogenicity.
• Previous studies showed variable predictive accuracy in carcinogenesis models based on mutagenic or non-mutagenic carcinogens.

Purpose of the Study:

• To evaluate the predictive accuracy of rat cancer SAR models for mutagens and non-mutagens.
• To assess the impact of carcinogen complexity (number of sites) on model performance.

Main Methods:

• Development and leave-one-out (LOO) validation of four rat cancer SAR models (Complete, Transgender, Male, Female).
• Analysis of concordance values for models trained on mutagenic versus non-mutagenic carcinogens.
• Comparison of SAR models based on the number of carcinogen sites (1-site, 2-site, 4+-site).

Main Results:

• LOO validation concordance values for the rat cancer SAR models ranged from 67% to 74%.
• Models accurately predicted mutagens (69-76% concordance) but poorly predicted non-mutagens (47-53% concordance).
• Increasing the number of carcinogen sites improved LOO concordance from 66% (1-site) to 79% (4+-site), with a higher proportion of mutagens.

Conclusions:

• Mutagenic chemicals significantly impact the accuracy of carcinogen SAR models.
• Accurate carcinogenicity prediction using SAR models may necessitate a two-step process, prioritizing mutagenicity assessment.
• The complexity of carcinogens, indicated by the number of sites, correlates with improved model performance and mutagenicity proportion.