Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

This article is part of the supplement: Proceedings of the Tenth Annual MCBIOS Conference

Open Access Proceedings

Differential reconstructed gene interaction networks for deriving toxicity threshold in chemical risk assessment

Yi Yang1, Andrew Maxwell1, Xiaowei Zhang2, Nan Wang1, Edward J Perkins3, Chaoyang Zhang1* and Ping Gong4*

Author Affiliations

1 School of Computing, University of Southern Mississippi, Hattiesburg, MS 39406, USA

2 State Key Laboratory of Pollution Control and Resource Reuse & School of the Environment, Nanjing University, Nanjing, China

3 Environmental Laboratory, US Army Engineer Research and Development Center, Vicksburg, MS 39180, USA

4 Badger Technical Services, LLC, San Antonio, TX 78216, USA

For all author emails, please log on.

BMC Bioinformatics 2013, 14(Suppl 14):S3  doi:10.1186/1471-2105-14-S14-S3

Published: 9 October 2013

Abstract

Background

Pathway alterations reflected as changes in gene expression regulation and gene interaction can result from cellular exposure to toxicants. Such information is often used to elucidate toxicological modes of action. From a risk assessment perspective, alterations in biological pathways are a rich resource for setting toxicant thresholds, which may be more sensitive and mechanism-informed than traditional toxicity endpoints. Here we developed a novel differential networks (DNs) approach to connect pathway perturbation with toxicity threshold setting.

Methods

Our DNs approach consists of 6 steps: time-series gene expression data collection, identification of altered genes, gene interaction network reconstruction, differential edge inference, mapping of genes with differential edges to pathways, and establishment of causal relationships between chemical concentration and perturbed pathways. A one-sample Gaussian process model and a linear regression model were used to identify genes that exhibited significant profile changes across an entire time course and between treatments, respectively. Interaction networks of differentially expressed (DE) genes were reconstructed for different treatments using a state space model and then compared to infer differential edges/interactions. DE genes possessing differential edges were mapped to biological pathways in databases such as KEGG pathways.

Results

Using the DNs approach, we analyzed a time-series Escherichia coli live cell gene expression dataset consisting of 4 treatments (control, 10, 100, 1000 mg/L naphthenic acids, NAs) and 18 time points. Through comparison of reconstructed networks and construction of differential networks, 80 genes were identified as DE genes with a significant number of differential edges, and 22 KEGG pathways were altered in a concentration-dependent manner. Some of these pathways were perturbed to a degree as high as 70% even at the lowest exposure concentration, implying a high sensitivity of our DNs approach.

Conclusions

Findings from this proof-of-concept study suggest that our approach has a great potential in providing a novel and sensitive tool for threshold setting in chemical risk assessment. In future work, we plan to analyze more time-series datasets with a full spectrum of concentrations and sufficient replications per treatment. The pathway alteration-derived thresholds will also be compared with those derived from apical endpoints such as cell growth rate.