The computational prediction of genotoxicity.

Area of Science:

• Computational toxicology
• Drug discovery
• Chemical safety assessment

Background:

• Computational prediction of genotoxicity is crucial for identifying potential human carcinogens early in development.
• Early identification of genotoxic compounds prevents progression of unsafe chemicals.

Purpose of the Study:

• To review key scientific advancements in predicting Ames mutagenicity and in vitro chromosome damage.
• To evaluate the performance and limitations of computational approaches in genotoxicity prediction.
• To discuss the application of these methods in modern drug discovery.

Main Methods:

• Review of scientific literature on computational genotoxicity prediction over the last 4-5 years.
• Analysis of validation exercises for computational models.
• Discussion of in silico system applicability in drug discovery.

Main Results:

• In silico systems demonstrate strong performance on public mutagenicity data.
• Predictive performance significantly decreases outside the established applicability domain.
• Key limitations include a lack of mechanistic structure-activity relationships and restricted access to high-quality proprietary data.

Conclusions:

• Current computational genotoxicity tools are effective within their data domain but limited beyond it.
• Advancements in predictive toxicology are hindered by insufficient mechanistic understanding and data accessibility.
• Future progress requires improved mechanistic insights and broader access to diverse, high-quality datasets.