DEXSeq pairwise group comparison

Question

[alex]

Hi,

Let's suppose I have RNA-seq data with two condition (Ctrl, Tr) measured at two time-points (12h, 24h), all having 4 replicates, so 16 samples in total. I am interested in looking at the exon usage with DEXSeq. Ideally, I want to do check for differential exon usage for 4 pair-wise comparisons, i) Ctrl.vs.Tr at 12h, ii) Ctrl. vs Tr at 24h, iii) Ctrl at 12h vs. Ctrl. at 24h and iv) Tr at 12h .vs Tr at 24h.

I've followed DEXSeq tutorial but I am not sure whether I understand how to use reduced and fullModel in this particular case. A simple approach seem to be to run the analyses separately, only keeping two groups at the time, but there are other posts out there mentioning that it is better to estimate dispersion on full data.

What would be best way forward? If keeping all the the data, how do I define the models for the comparison?

Would appreciate any help!

Answer 1

[BioRyder]

Hello,

In your case, make four groups [ i) Ctrl.vs.Tr at 12h, ii) Ctrl. vs Tr at 24h, iii) Ctrl at 12h vs. Ctrl. at 24h and iv ) Tr at 12h .vs Tr at 24h ] and do Deseq for each set. It will easily help you to see the differential expression of each set.

[alex]

Thanks, I'm leaning to this simple approach as well. I am still curious though whether it is possible to include all the data for dispersion estimation, and if so, how the pairwise comparisons should be then defined. d

[BioRyder]

It is possible to add all the data in dispersion estimation. Refer the following link. Hope it will clear your problem

http://cartwrightlab.wikispaces.com/DESeq

[alex]

Thanks, I know how to test for differentially expressed genes using DESeq or edgeR and how to define the group comparisons I want. This seems to be a bit different for differential exon usage in DEXSeq?