[BioC] edgeR: Identifying genes up regulated in one of multilple samples, and those stable across rsamples

[Gordon K Smyth]

Dear Dennis,

To find transcripts specific to one tissue versus the others, you need to 
test that tissue vs the others, but I'm sure you already know that.  If 
you have replicates of each tissue, you can do pairwise comparisons 
between the tissues, and choose transcripts which are consistently up 
in one vs all the others.  If you don't have replicates, there's no 
alternative really other than to treat the other 7 as a group.

Including lots of isoforms and using tags multiple times will multiple the 
number of rows in your data and therefore increase the amount of multiple 
testing.  Other than that, it is not a serious problem.  You could 
consider doing a gene-level analysis first, then repeating with an isoform 
level analysis for the significant genes found in the first step.

Best wishes
Gordon

> Date: Fri, 10 Jun 2011 16:49:05 +1000
> From: "Dennis Gascoigne" <d.gascoigne at imb.uq.edu.au>
> To: <bioconductor at r-project.org>
> Subject: [BioC] edgeR: Identifying genes up regulated in one of
> 	multiple	samples, and those stable across samples
>
> Hi;
>
> I have a list of transcripts with tag counts for each of them in 8 different
> tissues. My list includes many isoforms of similar genes etc. which is not a
> problem for me - but will it effect the stats from edger as obviously some
> tags are counted multiple times. There are 40M tags in each experiment so I
> am assuming the aggregate effect should prevent any great problems.
>
> My main questions are;
>
> 1.       I would like to identify which transcripts appear to be specific to
> an individual tissue from my 8
>
> 2.       Which genes are ubiquitously expressed across all 8
>
> Cheers
> Dennis

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}