very short 'time series' bulk rna-seq, testing two conditions

Question

[joha]

Hello, I have bulk rna-seq data that consist of two conditions (control, treated) tested across three time points (day 0, 1 and 3). We already observed by simple pair-wise (intra-condition) EdgeR that a big number of genes exhibit a change in expression between each two adjacent time points (day 1 vs 0 , then day 3 vs 1). We want to find significant differences INTER-conditions i.e genes whose change is significantly different in Treated vs Control at day 1 vs 0 , then day 3 vs 1. How could I do this with EdgeR? I have run EdgeR before but I do not know how to test both influence of treatment and time in one single design formula, and make sure that adjacent time points be tested (and NOT 3 vs 0 for exemple). I have 3 biological replicates by time-point and condition. Thanks for any help!

Answer 1

[ATpoint]

You could use a full-factorial design by combining the time variable and treatment variable like (simplified exampe):

sample  treatment   time    factorial

sample1 case    t0  case_t0
sample2 control t0  control_t0
sample3 case    t1  case_t1
sample4 control t1  control_t1
sample5 case    t3  case_t3
sample6 control t3  control_t3

and then design as ~factorial followed by making the desired contrasts. Contrasts could be something like:

(case_t1 - control_t1) - (case_t0 - control_t0)

That would test for genes where the difference between case and control is differential between time points. For example a modest change between case and control at t0 that is much greater in t1 would could come out as significant. This is called an interaction. See e.g. section 3.3 in the manual for more details and examples, https://www.bioconductor.org/packages/release/bioc/vignettes/edgeR/inst/doc/edgeRUsersGuide.pdf

[joha]

Thank you very much, your answer is exactly what we need. I apologize for my late reply, we were testing the interaction in edgeR as you kindly suggested.

[joha]

Thank you again, I didn't know I had to use "comment" to reply. Just a very last question, in DESeq2 is this also possible ?? (not very clear for me in DESeq2 vignette :( )

[ATpoint]

http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#interactions