Question: edgeR with unequal no.of replicate
0
3.9 years ago by
Nashik
Sushant Pawar0 wrote:

Hi all,

I am using edgeR for Gene expression analysis. My question is about unequal no.of replicate. If we have 4 replicate of Control and 1 Case Sample then is it fair to do analysis or we should go with edgeR without replicate like 1 case and 1 Control.

edger replicate • 961 views
modified 3.1 years ago by nancyjwahl70 • written 3.9 years ago by Sushant Pawar0
Answer: edgeR with unequal no.of replicate
2
3.9 years ago by
Aaron Lun25k
Cambridge, United Kingdom
Aaron Lun25k wrote:

edgeR can handle different numbers of replicates without any problems. However, it's generally better to have multiple replicates in each group as this gives more power to detect DE. With your current design, edgeR will (correctly) be more cautious when calling DE between groups, because high/low expression in the single Case sample might just be due to chance. If you had more Case replicates, edgeR could be more confident.

In any case, any replication is better than no replication at all. If you just use one Case against one Control, you won't have any information to estimate the dispersion. This would be the worst-case scenario, as you'd have to use an arbitrary dispersion parameter rather than something that's based on experimental data.

Answer: edgeR with unequal no.of replicate
0
3.1 years ago by
nancyjwahl70 wrote:

Hi,

I'm glad to hear that edgeR can handle different numbers of replicates. I have the following situation:

I have 2 factors in my RNA-Seq experiment, 1. differences in maturity of my maize kernels, which very clearly result in differential expression and 2. differences in inoculation treatments with fungal spores (or no inoculation), which have a less pronounced effect. At most, when I look at differences due to treatments within a maturity group, I only have 2 replicates. I was going to eliminate the one maturity group that had no treatment replicates until I saw many defense genes highly expressed within that group, like no other. So now I would like to include that as well so I can compare inoculation effect vs no inoculation in that maturity group. I have chosen edgeR as my analysis package. Do I need to be concerned about estimating a dispersion factor if I run all the samples together? Does edgeR draw from the other replicated samples and apply the same factor to all? It appears that way from statements in the manual. Thanks!

Nancy