An Investigation Of Gene Networks Influenced By Low Dose Ionizing Radiation Using Statistical And Graph Theoretical Algorithms

Increased application of radiation in health and security sectors has raised concerns about its deleterious effects. Ionizing radiation (IR) less than 10cGys is considered low dose ionizing radiation (LDIR) by the National Research Committee to assess health risks from exposure to low levels of IR.

It is hard to extract the effects of mild stimulus such as LDIR on gene expression profiles using simple differential expression. We hypothesized that differential correlation instead would capture the effects of LDIR on mutual relationships between genes. We tested this hypothesis on expression profiles from five inbred strains of mice treated with LDIR. Whereas ANOVA detected little effect of LDIR on gene expression, a differential correlation graph generated by a two stage statistical filter revealed gene networks enriched with genes implicated in radiation response, DNA damage repair, apoptosis, cancer and immune system.

To mimic the effects of radiation on human populations, we profiled baseline expression of recombinant inbred strains of BXD mice derived from a cross between C57BL/6J and DBA/2J standard inbred strains. To establish a threshold for extraction of gene networks from the baseline expression profiles, we compared gene enrichment in paracliques obtained at different absolute Pearson correlations (APC) using graph algorithms. Gene networks extracted at statistically significant APC (r≈0.41) exhibited even better enrichment of genes participating in common biological processes than networks extracted at higher APCs from 0.6 to 0.875.

Since immune response is influenced by LDIR, we investigated the effects of genetic background on variability of immune system in a population of BXD mice. Considering immune response as a complex trait, we identified significant QTLs explaining the ratio of CD8+ and CD4+ T-cells. Multiple regression modeling of genes neighboring statistically significant QTLs identified three candidate genes (Ptprk,Acp1 and Lamb1-1) explaining 61% variance of ratio of CD4+ and CD8+ T cells. Expression profiling of parental strains of BXD mice also revealed effects of LDIR and LDIR*strain on expression of genes related to immune response. Thus using an integrated approach involving transcriptomic, SNP and immunological data, we have developed novel methods to pinpoint candidate gene networks putatively influenced by LDIR.