Characterization of Cell Type Specific miRNA Profiles and Application to miRNA Profiles Derived from Tissues
Alexander S Baras, Matthew McCall, Toby Cornish, Marc Halushka. Johns Hopkins Hospital, Baltimore, MD; University of Rochester, Rochester, NY
Background: MicroRNAs (miRNAs) are highly conserved RNAs that serve as master regulators of gene expression. They are exciting biomarkers and therapeutic targets. Studies have found altered expression of miRNAs in tissue disease states; however, the interpretation of these studies may be fundamentally flawed because the changing cellular composition of tissues in disease states is not adequately accounted for. Techniques to control for this source of variability should increase the relevance of changes in miRNA species identified in profiling studies of diseased tissue.
Design: Our approach characterizes both factors that can affect miRNA expression levels: altered cellular composition of tissue and disease specific expression changes. We have utilized miRNA expression profiles from a panel of cell types to deconvolute observed miRNA profiles generated from various tissues. We have then explored the impact of this two factor analysis via a modeling study of ulcerative colitis.
Results: We utilized publicly available data from 16 cell types including inflammatory, endothelial, stromal, and epithelial cells to predict the cellular composition and model the observed expression of the miRNA expression profile from 4 tissues (Fig. 1A). The predicted cellular compositions are consistent with the expected composition of these tissues and the modeled miRNA levels are highly correlated to the observed signals. We next performed a modeling experiment of ulcerative colitis using image analysis to predict cellular ratio changes (Fig. 1B). We generated hypothetical miRNA profiling data of normal colon and ulcerative colitis in which we manually altered only 3 miRNAs at the cell level – potential disease specific changes. We then analyzed the data via routine univariate hypothesis testing with and without correction for changes in tissue composition. Without tissue composition correction, most miRNAs identified are false positives (Fig. 1C).
Conclusions: The findings highlight the importance of understanding the cellular contribution of miRNAs to tissue expression data and controlling for changes in cellular composition in disease analysis.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Tuesday, March 5, 2013 9:30 AM
Poster Session III # 238, Tuesday Morning