Literature-based compound profiling: application to toxicogenomics.

Area of Science:

• Computational toxicology
• Bioinformatics
• Drug discovery and development

Background:

• Rising drug development costs necessitate early detection of adverse drug effects.
• Gene expression profiling, including Gene Ontology (GO) and pathway analyses, is used for toxicity assessment.
• Existing methods require enhancement for comprehensive toxicity and mode of action evaluation.

Purpose of the Study:

• To introduce a novel approach for compound profiling using literature-derived information.
• To evaluate compound toxicity and elucidate the mode of toxicity.
• To establish a method for early detection of adverse drug effects.

Main Methods:

• Text mining of Medline abstracts to build gene annotations linking genes, pathology terms, biological processes, and pathways.
• Generation of compound-specific keyword fingerprints based on over-represented keywords in regulated genes post-compound administration.
• Analysis of microarray data from rat liver treated with 11 hepatotoxicants to validate keyword fingerprinting.

Main Results:

• Each compound generated a specific keyword fingerprint correlating with observed histopathological events, unlike random gene sets.
• Analysis of keyword profiles for specific compounds revealed distinct modes of action.
• Literature network visualization enabled construction of a mode of toxicity proposal for methapyrilene (MPy).

Conclusions:

• Compound keyword fingerprinting, leveraging literature data, is a powerful tool for compound profiling.
• This approach enables effective evaluation of compound toxicity and detailed analysis of the mode of action.
• The method supports early identification of potential drug hazards in the development pipeline.