[BioC] limma-voom

[Steve Lianoglou]

Hi,

Just wanted to comment on the "paired" TCGA data:

On Wed, Feb 5, 2014 at 2:38 PM, Gordon K Smyth <smyth at wehi.edu.au> wrote:
[snip]
> I have analysed similar data from TCGA, and the separation is really much
> too good.  Furthermore there is little evidence of matching.  It would
> appear that the matched normal cells are not really comparable cells to the
> tumours.  The matched normal samples may actually be profiles of whole
> blood, a cell type which has a radically different expression profile to
> epithelial cells.  I really wonder what one can hope to learn about cancer
> by comparing epithelial tumours to blood.

There are samples where you have the "correct" matched sample for gene
expression, as indeed comparing a solid tumor to "normal blood" to
identify changes in gene expression would make very little sense.

You just need to be careful in parsing the sample barcodes to ensure
you are working with the correct stuff.

An explanation of the barcode scheme is here:
https://wiki.nci.nih.gov/display/TCGA/TCGA+Barcode

You are interested in decoding the "sample" value, which ranges from 01-29.

Explanation is here:
https://tcga-data.nci.nih.gov/datareports/codeTablesReport.htm?codeTable=Sample%20type

There are samples taken from "normal blood" (code = 10), however
code=11 are solid tissue normals (presumably from the correct tissue
;-)

The preponderance of blood samples is, I believe, primarily to be used
in order to identify somatic mutations -- presumably using the
exome-seq from "normal blood" would be cleaner than primary tissue,
depending on how close to the tumor the tissue was taken from.

To figure out what was extracted from the "sample" (DNA, RNA, etc),
you just decode the "Analyte" part of the barcode:
https://tcga-data.nci.nih.gov/datareports/codeTablesReport.htm?codeTable=portion%20analyte


HTH,
-steve

-- 
Steve Lianoglou
Computational Biologist
Genentech