Background: Identification of cancer hazards is the first step towards cancer prevention. The IARC Monographs Programme has evaluated nearly 1000 agents for carcinogenic potential since 1971. A chemoinformatics approach can be used to systematize, inform and further increase efficiency in selecting agents for evaluation. IARC will evaluate several pesticides in February 2015. Considering the large number of chemically similar pesticides, a ranking method is needed to determine their priority for evaluation.
Aim:Integration of a chemoinformatics and computational approach for ranking of pesticide chemicals for the IARC evaluation process.
Methods:Information from USEPA pesticide database, PubChem BioAssay, ToxRefDB, PubChem Compound DB, NCBI BioSytems and NCBI Pubmed databases were integrated using web technologies, chemoinformatics algorithms and network graphs to develop the ranking software.
Results:A total 5700 entries from USEPA pesticide structure database were downloaded. Up to 3100 entries had been associated with PubChem compound identifiers. Use of network clustering algorithms on chemical similarity maps suggested up to 40 distinct chemical clusters of pesticides can be obtained, representing the vast chemical diversity among pesticides. Overlaying the retrieved information from various databases on these maps identified clusters of pesticides that can be given high priority in the evaluation process. Pesticides already evaluated by IARC monographs were ranked high. Comparison of these maps highlighted the clusters of pesticides that have been studied in-vitro and in-animal but not in epidemiological studies. A web interface to access the enriched maps will be provided online.
Conclusions:Use of chemical clustering and bioassay and literature data yielded cluster level ranking of pesticides for their evaluation of carcinogenic potential. The approach can readily be extended to other compounds classes such as drugs, environmental pollutants, endogenous metabolites and food components. Furthermore, untargeted metabolomics data in prospective cohorts can be screened for several pesticides that have evidence from only animal or in-vitro studies.