|
Abstract:
|
In order to develop targeted cancer treatment strategies, tissue samples from tumors are evaluated to identify biomarkers. Using RNA transcript quantification to determine gene expression profile, mixed tissue samples must be analyzed to determine underlying component tissues, and individual tissue-specific expression profiles. In this way the unique expression profile of the cancer is identified, eliminating background from other healthy tissue. A Bayesian method for mixed tissue expression analysis has been developed previously by Ekkila et al [DSection. Bioinformatics 2010] utilizing microarray expression data and prior information on constituent tissue proportion. This work presents an adaptation of the approach of Erkkila et al to RNA sequencing data, as this method of transcript quantification has advantages over microarrays. However microarray data are continuous measures of gene probes and RNA-seq data are discrete counts mapped to genes, so RNA-seq data must be pre-processed before use with programs developed for microarrays. The effects of using this new data type with the model will also be evaluated, specifically with regard to convergence and computational efficiency.
|
