Question: How to analyze RNAseq data with two/multiple treatment groups and each subject has pre and post-treatment measures
0
gravatar for wwu222
9 weeks ago by
wwu2220
wwu2220 wrote:

Hi,

I am working on a set of RNAseq data that has the following design:

  1. patients are randomized to different treatment groups: control, trt1 and trt2
  2. each patient has pre and post treatment RNAseq data: time0 , time1

I am interested in getting treatment effect (not time effect). The model mentioned in DESeq2, edgeR and Limma helps to get time effect which is not the interest.

Any suggestion how to get treatment effect for RNAseq data with this design?

Thank you very much!

rna-seq • 166 views
ADD COMMENTlink modified 8 weeks ago by Kevin Blighe42k • written 9 weeks ago by wwu2220
1
gravatar for Kevin Blighe
8 weeks ago by
Kevin Blighe42k
Guy's Hospital, London
Kevin Blighe42k wrote:

To get treatment effect across time, you can perform ANOVA (LRT - likelihood ratio test):

dds <- DESeqDataSetFromMatrix(..., design = ~ treatment + time + treatment:time)
dds <- DESeq(dds, test="LRT", reduced = ~ treatment + time)
res <- results(dds)

If you simply want to 'ignore' the time effect and just control for it, i.e., as a covariate, then consider simply doing ~ treatment + time, followed by pairwise comparisons between your treatment levels.

ADD COMMENTlink modified 8 weeks ago • written 8 weeks ago by Kevin Blighe42k

Hi Kevin,

Thank you very much for your timely reply and great suggestions. I think the second suggestion is the answer I am looking for. However, each patient has pre, post paired design, I am wondering whether we have to consider this in the model?

Thank you very much!

ADD REPLYlink written 8 weeks ago by wwu2220

So, you want to adjust for the patient-specific effects, too? - then you can include PatientID as a covariate.

ADD REPLYlink written 8 weeks ago by Kevin Blighe42k

Hi Kevin, really appreciate your suggestion! I will try it out. Thank you very much!

ADD REPLYlink written 8 weeks ago by wwu2220

Cool. The model may then simply be ~ PatientID + treatment. This model is then effectively a paired analysis and looking at the effect of treatment. Time is captured to some level via PatientID. These model formulae can be somewhat confusing and you should always get multiple opinions.

ADD REPLYlink written 8 weeks ago by Kevin Blighe42k
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 652 users visited in the last hour