[BioC] DESeq - skewed MAplots and other problems

[Michal Lulu]

Hi,


At first I'd like to explain that my RNAseq experiments involve RNAi
depletions of RNA decay factors. The libraries are ribodepleted, paired end
and stranded. After mapping with tophat I use HTSeq/DESeq combo to discover
DE genes (among tophat genes.gtf)

Using raw counts I run DESeq on default settings and I get certain numer of
signif. DE genes, approx. 1:1 up- to down-regulated. Everything seems fine
though my MA plots are a bit skewed (example attached), there is a clear
slope suggesting that more upregulation of genes of lower expression.
Should I worry about this?

I also tried to compare DESeq normalization with normalization to spike-ins
present in the libraries, but the size factors assigned by DESeq seems
much more accurate; although it's unclear why ?


Finally, I turned to normalization that is not recommended by authors but
some people do this: http://jura.wi.mit.edu/bio/education/hot_topics/rnaseq/
*rnaseqde_dec2011*.*pdf*

Strangely, when I introduce pseudocounts, which in principal should not
affect analyzes that much, they actually do. New upregulated hits appear
and most downregulations disappear, importantly most of the upregulated
enriched already existing clusters in GO. This suggests they may be real.


Please note this is my second RNAseq analysis, so I'm really fresh and I
would appreciate a lay explanation :)


Cheers,

Michael
-------------- next part --------------
A non-text attachment was scrubbed...
Name: analysis.pdf
Type: application/pdf
Size: 260615 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/bioconductor/attachments/20121004/18808c40/attachment.pdf>