Related Experiment Videos
A simple and readily integratable approach to toxicity prediction.
Steven M Muskal1, Sanjiv Kumar Jha, M Phani Kishore
1Sertanty, Inc, 1735 N. First St., Suite 102, San Jose, California 95112, USA. smuskal@sertanty.com
Journal of Chemical Information and Computer Sciences
|September 23, 2003
Summary
A new computational method predicts compound toxicity by averaging data from similar structures. This approach shows strong predictive power for acute toxicity, with potential for broader applications in chemical safety assessment.
Area of Science:
- Computational toxicology
- cheminformatics
- Drug discovery and development
Background:
- Assessing chemical compound toxicity is crucial for safety and drug development.
- Existing methods can be time-consuming and resource-intensive.
- Predictive toxicology models aim to estimate toxicity computationally.
Purpose of the Study:
- To develop a computational strategy for assessing compound toxicity.
- To leverage structural similarity for toxicity prediction.
- To evaluate the strategy's predictive performance using established metrics.
Main Methods:
- Developed a computational approach based on the toxicity of structurally similar compounds.
- Utilized a reference dataset of 13,645 compounds with oral, rat-LD(50) data.
- Performed leave-one-out cross-validation simulations to assess predictive accuracy (q²).
Main Results:
- The strategy achieved a predictive correlation (q²) of 0.82, predicting 57.3% of compounds in the reference set.
- Applied to a set of 1,781 drugs, the method yielded a q² of 0.74, predicting 51.8% of compounds.
- Increasing reference set size significantly improved prediction quality and quantity.
Conclusions:
- The developed computational strategy effectively predicts acute compound toxicity based on structural similarity.
- The approach is highly extensible and adaptable for various toxicity endpoints.
- Further application to subchronic and chronic toxicity data is feasible with adequate reference datasets.