Application of computational systems biology to explore environmental toxicity hazards

Abstract

Background: Computer-based modeling is part of a new approach to predictive toxicology.

Objectives: We investigated the usefulness of an integrated computational systems biology approach in a case study involving the isomers and metabolites of the pesticide dichlorodiphenyltrichloroethane (DDT) to ascertain their possible links to relevant adverse effects.

Methods: We extracted chemical-protein association networks for each DDT isomer and its metabolites using ChemProt, a disease chemical biology database that includes both binding and gene expression data, and we explored protein-protein interactions using a human interactome network. To identify associated dysfunctions and diseases, we integrated protein-disease annotations into the protein complexes using the Online Mendelian Inheritance in Man database and the Comparative Toxicogenomics Database.

Results: We found 175 human proteins linked to p,p'-DDT, and 187 to o,p'-DDT.Dichlorodiphenyldichloroethylene (p,p'-DDE) was the metabolite with the highest number of links, with 52. We grouped proteins for each compound based on their disease annotations. Although the two data sources differed in linkage to diseases, integrated results predicted that most diseases were linked to the two DDT isomers. Asthma was uniquely linked with p,p'-DDT, and autism with o,p'-DDT. Several reproductive and neurobehavioral outcomes and cancer types were linked to all three compounds.

Conclusions: Computer-based modeling relies on available information. Although differences in linkages to proteins may be due to incomplete data, our results appear meaningful and suggest that the parent DDT compounds may be responsible for more disease connections than the metabolites. The findings illustrate the potential use of computational approaches to toxicology.

Figure 1

Overview of the systems chemical biology three-step approach. (1) Extraction of existing knowledge using a disease chemical biology database (ChemProt) to generate chemical–protein networks for p,p´-DDT. (2) Creation of protein complexes by protein enrichment using a high-confidence set of experimental protein–protein interactions. (3) Statistical ranking of diseases (D) known or predicted to be linked to p,p´-DDT after integration of protein–disease annotations to protein complexes based on information in the Online Mendelian Inheritance in Man (OMIM) database and the Comparative Toxicogenomics Database (CTD).

Figure 2

Venn diagram showing the number of diseases overlapping between p,p´-DDE and p,p´-DDT, using disease annotations extracted from the OMIM and the CTD databases. Data on o,p´-DDT are not shown, as most of its disease links overlapped with p,p´-DDT.

Figure 3

Disease–chemical associations network. The circles represent diseases, with colors representing phenotype categories: red, reproductive disorders; blue, neurodevelopmental-related diseases; green, cancers; gray, other diseases. Rectangles represent the three chemicals studied. The heavier the weight of the connecting lines, the greater the number of proteins linking a chemical to a disease (determined using the OMIM and CTD databases and ChemProt as resources).