Question

Testing for differential expression variability with edgeR

0

Entering edit mode

Miguel Gallach ▴ 50

@miguel-gallach-5128

Last seen 10.3 years ago

Dear Bioconductor list, I am using edgeR to test for differential expression but I also I would like to know how to test whether expression variability (per gene) is significantly different between biological groups or treatments. For instance, the next is my experimental design, according to which I have two treatments (Hot and Cold) and two biological groups (Hot Adapted and Cold adapted; two replicates each). $samples group lib.size norm.factors R4.Hot HotAdaptedHot 17409289 0.9881635 R5.Hot HotAdaptedHot 17642552 1.0818144 R9.Hot ColdAdaptedHot 20010974 0.8621807 R10.Hot ColdAdaptedHot 14064143 0.8932791 R4.Cold HotAdaptedCold 11968317 1.0061084 R5.Cold HotAdaptedCold 11072832 1.0523857 R9.Cold ColdAdaptedCold 22386103 1.0520949 R10.Cold ColdAdaptedCold 17408532 1.0903311 I would like to detect those genes for which the expression variability between R4.Hot and R5.Hot is significantly different to that obtained for R4.Cold and R5.Cold. Can I use the tagwise dispersion values provided by edgeR, and how?? Many thanks! Miguel [[alternative HTML version deleted]]

edgeR edgeR • 992 views

ADD COMMENT • link updated 3.8 years ago by Gordon Smyth 52k • written 12.8 years ago by Miguel Gallach ▴ 50

score 0 · Answer 1 · 2012-02-23

Dear Miguel,

Well, you could obviously subset your data, and compute the tagwise dispersions separately for each subset, and compare the two. Doing a formal test is far more difficult and would be genewise anyway.

I find it hard to imagine in what context it would make sense to test whether two libraries (R4.Hot and R5.Hot) are more different than two other libraries. For a particular gene? Averaged over genes?

Could you not just make an MDS plot and look to see how far apart the samples are from each other?

What are you trying to achieve here from a biological point of view? What biological question are you trying to answer?

Best wishes
Gordon