Hi all,

I am new to DESeq2 analysis. I have RNASeq raw count data from three different time points namely 0h, 10h, 2d. Every time point has triplicates for Control samples, Treatment-A, TreatmentB, TreatmentC. All I want to find is the differential expressed genes between Treated against Control samples(Treatment-A vs Control; Treatment-B vs Control; Treatment-C vs Control) for every time point. Now I would like to know whether , I should load raw counts from all the samples together in every time point into R and compute diff. expression analysis OR load the raw counts of particular time point(i.e. Control_0h, Treatment-A, Treatment-B and Treatment-C) at a time and perform differential gene expression analysis.

Can anyone guide me. Thanks in advance.

Regards

Chudar

I'd recommend putting all the samples into one DESeqDataSet, and using a design of ~ condition + time + condition:time

There is an example of a time course analysis in our workflow (just for one treatment though):

www.bioconductor.org/help/workflows/rnaseqGene/#time

The likelihood ratio test described in that workflow example for your case would give any genes which show a difference due to any of the treatments at any time point. This is not however what you asked about in your question. 

Instead, you can test single interaction terms like so:

res <- results(dds, name="condX.time10h", test="Wald")

By testing the interaction terms one by one, you can compare the condition X vs control effect at a given time point after time 0h, controlling for any difference that existed for that treatment compared to control at time 0h (this is usually what investigators are interested in).