Complex design with paired and unpaired samples with edgeR RNA-seq DE analysis
1
0
Entering edit mode
Basti ▴ 150
@7d45153c
Last seen 1 day ago
France

Hi everyone,

I have a complex RNA-seq experimental design like this :

condition subject
A         1
A         2
A         3
A         4
B         1
B         2
B         3
C         5
C         6
C         7
C         8


First my goal is to compare A vs C condition but I also want to compare A vs B, which have some paired samples.

In my case it seems that design=model.matrix(~subject+condition) will not work because my matrix is not of full rank.

Would you have any advice on how should I proceed to perform these 2 separate comparisons with paired and non-paired samples? Thanks in advance for enlightening me

edgeR RNASeq • 98 views
2
Entering edit mode
@gordon-smyth
Last seen 9 minutes ago
WEHI, Melbourne, Australia

In edgeR, you would need to do two separate analyses, a paired analysis of A vs B and an unpaired analysis of A vs C. That might be the best.

In limma, there is the extra possibility of a combined analysis using duplicateCorrelation to treat subject as a random effect. The only real advantage of that approach in this case would be to include subject 4 when comparing A vs B.