The analysis of gene coexpression is often used in order to create system level views of cellular circuitry. However, it is widely known that the experimental noise induced through expression heterogeneity has the potential to cause spurious correlations and to hinder the ability to find true correlations among functionally related genes.

We have developed a method to battle this problem called Mixed Model Coexpression (MMC). MMC works by defining the coexpression between two genes within a mixed effects framework. With this formulation our method is able to use the observed correlation structure between microarray experiments to correct for the effects of expression heterogeneity and calculate a value of coexpression that is not influenced by large scale confounding effects.

Find the paper at http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3117390/

Contact: Nick Furlotte (nfurlott at cs dot ucla dot edu)

Funding Information:
National Institute of Health training grant T32-HG002536 (to N.F.). National Science Foundation (No. 0513612, No. 0731455 and No. 0729049) (to N.F., H.M.K., C.Y., E.E.); and National Institutes of Health (1K25HL080079 and U01-DA024417); Samsung Scholarship, the National Human Genome Research Institute (Grants No. HG00521401 to H.M.K.); National Institute for Mental Health NIMH No. NH084698, and GlaxoSmithKline (in part). UCLA subcontract of contract N01-ES-45530 from the National Toxicology Program/National Institute of Environmental Health Sciences to Perlegen Sciences (in part).