Related Experiment Videos
Applications of computers to toxicological research
1Laboratory of Medicinal Chemistry, National Cancer Institute, National Institutes of Health, Bethesda, Maryland 20892.
Chemical Research in Toxicology
|November 1, 1993
Summary
Computers aid toxicology by managing data and predicting toxicity. While effective for obvious cases, computer predictions of carcinogenicity lag behind human experts, though ongoing development promises improvements.
Area of Science:
- Toxicology
- Computational Chemistry
- Regulatory Science
Background:
- Computers are integral to modern toxicology, facilitating data management and database searching.
- Applications extend to regulatory compliance, driven by cost-benefit analyses.
- Predictive toxicology using computational methods is a key area of development.
Purpose of the Study:
- To evaluate the efficacy of computer-based methods in toxicity prediction.
- To compare computational toxicology predictions against human expert performance.
- To identify limitations and potential improvements in predictive toxicology systems.
Main Methods:
- Utilized computational methods for toxicity property estimation.
- Compared computer predictions with human expert assessments in a head-to-head study.
- Analyzed performance based on the 'obviousness' of compound carcinogenicity.
Main Results:
- Computer methods achieved 49-59% accuracy in predicting carcinogenicity.
- Human experts demonstrated higher accuracy, scoring between 65% and 84%.
- Both humans and computers performed well on obvious cases, with humans showing better adaptability for complex scenarios.
Conclusions:
- Current computational toxicology methods show mixed results in real-world applications.
- Human experts currently outperform computers in predicting carcinogenicity, especially for non-obvious compounds.
- Continued development of computational models and learning sets is expected to improve predictive accuracy over time.