Introduction
Tumor samples consisting of both cancer and stromal tissues are a major cause of under-detection of gene expression biomarkers for cancer prognosis. In silico dissection of mixture components is essential for analyzing expression data generated in cancer studies. We propose a statistical approach for deconvolving mixed cancer transcriptomes (DeMix), that assesses tumor-specific proportions in mixed tumor samples using the raw-measured data and reconstitutes individual gene expressions of normal and tumor samples.
Reference
DeMix: Deconvolution for Mixed Cancer Transcriptomes Using
Raw Measured Data.
Ahn J, Yuan Y, Parmigiani G, Suraokar MB, Diao L, Wistuba II, and Wang W. Bioinformatics 2013 doi: 10.1093/bioinformatics/btt301. [Abstract] [PDF]
Supplemental materials
Version
Version 1.0.1: the Mac version, the Linux version.
This version is for the matched or unmathed samples without reference genes
No need to install. Zipped file includes following three files. Require R-3.0.0 or above.
demix.so : gcc compiled c function linked with R
TumorHv1.r : executable R function.
R_run.r : Example code