I am using a pipeline (bam -> featurecount-> EdgeR) to do some RNASeq analysis of several groups and sub-groups. For example, I have the following dataset with two types (T1 and T2) and T1 has three groups G1, G2 and G3, T2 has G5 and G6 groups. I would like to find TPM numbers, fold-change and FDR for all the genes and Groups.

Type Group SRA_Run#
T1   G1.   <some SRA ids e.g. SRR123456, SRR112233>
T1   G2.   <some SRA>
T1.  G3.   <some SRA>
T2.  G5.   <some SRA>
T2.  G6.   <some SRA>

I have individual BAM files generated for each SRA Run. I also individual read counts from featureCount for each SRA Run. I was not sure if I need to merge individual read counts for each group or run featureCount using all bam files of a group at once. Does it make any difference in the analysis that follows.

I also would like to know if I can run EdgeR at once for all the types and groups at once for doing differential gene expression for each group of Type 1 (T1) against every other group of Type 2 (T2). I am bit stuck at this point how to proceed. Any help would be appreciated.


1 Answer 1


It makes no difference if you process the BAM files one at a time with featureCounts or all together, except that it changes how you have to read the files into R.

You can supply edgeR with lists of contrasts to have it compute fold-changes and p-values for. Please have a look at the edgeR user guide for examples.

  • $\begingroup$ Thanks @Devon !! So, I can simply merge the count files, right? Also, I am taking fastq from different SRA runs, so they can be considered as replicates? I am asking because I have some dataset that contains 10 SRA runs for one sample and 4 SRA runs for another sample. $\endgroup$
    – SBDK8219
    Commented Mar 23, 2021 at 14:13
  • $\begingroup$ It depends on whether they're runs of the same sample or not. If they're from the same sample then they're not replicates (they can be merged together). $\endgroup$
    – Devon Ryan
    Commented Mar 23, 2021 at 14:46
  • $\begingroup$ Thanks. Yes, they are from the same sample. In my original question, group G1 is the name of the sample and it has few SRA runs. Just one more question, will it be an issue if I use different number of SRA runs for two different groups? Also, one pipeline uses RUVr for normalizing the counts, so is it advised to run RUVr for merged featureCount results before running EdgeR? Please pardon my lack of knowledge about RNA-Seq analysis as I am bit new to this. $\endgroup$
    – SBDK8219
    Commented Mar 23, 2021 at 15:15
  • $\begingroup$ EdgeR and similar packages will handle the library size normalization. Nothing should be done on the individual runs before merging them. $\endgroup$
    – Devon Ryan
    Commented Mar 24, 2021 at 8:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.