Computational techniques for the prediction of toxicity.

Area of Science:

• Computational toxicology
• Cheminformatics
• Structure-activity relationships

Background:

• Computational methods for predicting biological activity have existed since the 1960s.
• Early applications in toxicology prediction faced criticism.
• This review examines three distinct toxicity prediction approaches.

Purpose of the Study:

• To review existing computational methods for toxicity prediction.
• To evaluate a new system, APEX, for identifying toxicophores.
• To assess the performance of APEX in predicting mutagenicity.

Main Methods:

• Review of three computational toxicity prediction strategies.
• Application of the APEX system to a literature dataset of mutagenicity.
• Training and testing the APEX system on a dataset of 105 compounds.

Main Results:

• The APEX system achieved a 75% success rate in predicting mutagenicity.
• Comparable prediction systems, trained on larger datasets, reported 85-90% accuracy.
• The study highlights the potential and limitations of current computational toxicology tools.

Conclusions:

• Computational toxicology offers valuable tools for predicting toxicological endpoints.
• Newer systems like APEX show moderate success but require further development.
• Ongoing advancements in computational methods are crucial for improving toxicity prediction accuracy.