[Bioc-sig-seq] EdgeR Questions regarding prior.n tagwise dispersion

Mads Sønderkær mson at bio.aau.dk
Mon May 30 08:56:21 CEST 2011


Hi,
Introduction:
I'm analyzing a high replicate tag based mRNAseq (SAGE like) data set. In the experiment I have 2x 47 replicates of true biological samples of potato leaves. I also have triplicates of libraries originating from the same crushed leaf material, but from different library preps. I hope to be able to evaluate biological and technical variation from these. I have chosen to use EgdeR for the analysis, while it has some very attractive features. One is that it is possible to estimate the dispersion.
For now I can conclude that the tagwise dispersion model fits the data better than the poisson and the common dispersion model (not suerpringly).

Howerver when calculating the tagwise dispersion, prior.n and used.prop needed to be defined. I have looked in the forum and found 1 E-mail regarding prior.n.
My question is:
Is there a way to find the optimal value for prior.n - and how is this evaluated? I'm thinking of subtracting the raw variance from the calculated tagwise variance, and then look at the absolute SUM (the sum of error). Is this a good way to go?

Regarding the d$tagwise.dispersion (here d is a DGEList after estimateTagwiseDisp). I believe the object is a vector, but there are no row names. My question is now if the order of  d$tagwise.dispersion is the same as rownames(d$counts)?

Session info:
R version 2.13.0 (2011-04-13)
Platform: x86_64-pc-mingw32/x64 (64-bit)

other attached packages:
[1] edgeR_2.2.5

Kind regards
Mads Sønderkær

[cid:image001.jpg at 01CC1EA5.94AFAD40]

Mads Sønderkær
PhD Student
Mobile: +45 3053 2492

Aalborg University
Department of Biotechnology, Chemistry and Environmental Engineering
Sohngaardsholmsvej 49
Room 17
9000 Aalborg
Denmark
Phone: +45 9940 8461
www.bio.aau.dk<http://www.bio.aau.dk/>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://stat.ethz.ch/pipermail/bioc-sig-sequencing/attachments/20110530/88d6669d/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.jpg
Type: image/jpeg
Size: 2195 bytes
Desc: image001.jpg
URL: <https://stat.ethz.ch/pipermail/bioc-sig-sequencing/attachments/20110530/88d6669d/attachment-0001.jpg>


More information about the Bioc-sig-sequencing mailing list