EdgeR Multivariate Analysis question
2
0
Entering edit mode
@joangibert14-13723
Last seen 6.5 years ago

Hi!

I was wondering... how do you specify edgeR that you want to obtain only the DE genes in Treatment B compared to Treatment A and Controls. I thought about using the mean of Control and Treatment A like:

TreatmentBVsControlandTA = makeContrasts(TreatmentB-(Control+TreatmentA)/2, levels=design)

But I guess that I would loose tones of genes due to possible high deviation between Control and Treatment A. Any idea how to overcome this issue?
Thanks!

edger limma • 584 views
ADD COMMENT
2
Entering edit mode
Aaron Lun ★ 28k
@alun
Last seen 46 minutes ago
The city by the bay

First you need to decide what you mean by "Treatment B compared to Treatment A and Controls". If you want to find genes that are DE in B vs A and in B vs Controls, then you should do two separate contrasts and intersect the results (presumably also filtering on consistent sign of the log-fold change).

Your proposed approach is to find genes that are DE between B and the average of the A and Control groups. This is also an acceptable strategy. It's a more relaxed approach than taking the intersection, and so will yield more DE genes; but at the cost of interpretability, as you don't know whether the genes from this contrast are DE (or even in the same direction) against both of A and control.

Note that deviation between Control and A is not directly relevant to power. The dispersion is still being estimated using your full design matrix, edgeR is not treating Control and A as replicates of one another. So you won't necessarily "lose tons of genes" from using this approach.

ADD COMMENT
0
Entering edit mode
@gordon-smyth
Last seen 22 minutes ago
WEHI, Melbourne, Australia

Just following up Aaron's answer, the contrast you have given is a very powerful way of finding genes DE in Treatment B compared to the average of A and Controls. You will definitely not lose genes that you would have wanted to detect because of deviations between Control and Treatment.

ADD COMMENT

Login before adding your answer.

Traffic: 625 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6