This is a test version of Biostars. For the public version, visit https://www.biostars.org.
RNA Seq on multiple samples

Hi. I want to compare samples from 4 different strains. For each strain I have 3 biological and 3 technical samples. What is the best way to go using cufflinks?

Now this is what I do:

  1. tophat on all samples.
  2. merge the bam files of the technical and the biological sample
  3. run cufflinks on the merged.bam file for each sample
  4. merge the transcripts.gtf file from the 3 samples in each strain
  5. perform cuffquant with the merged.gtf file
  6. run cuffnorm on the samples of each strain

I repeat this for all the strains that I have

Ideally I want to compare the expression of all the strains together (not pairwise comparisons).

Let me know if you have any idea of what is the best way to go

rna-seq next-gen-sequencing alignment

1 answer

Tuxedo cannot handle technical replicates. If I were you, I'd consider the following route:

Gene Level: Align with your favourite aligner -> count using htSeq_Count or rSubRead -> Use DESeq2 for Differential Gene Expression. (read about technical replicates here in section 1.3.7 and how to collapse them).

Transcript Level: Transcriptome quantification using Salmon (read about making Salmon counts friendly for Sleuth here) or Kallisto -> Differential Transcript Expression using Sluth (I haven't used technical replicates here, so you may have to do some reading, but it's by far the best way to do differential transcript expression)

edit: To answer your original question, your way of accounting for technical replicates seems right, however I'd consider feeding your cxb files from cuffQuant, into CuffDiff.

edit edit: If you're looking for a more flexible way to model your experiment, besides pairwise differential expression, then the routes I explained above are your best shot.

I also performed cuffdiff but it does pairwise comparisons but as far as I know DESeq does pairwise comparisons as well. Am I right?

DESeq2 allows you to model your experiment and derive your contrast of interest. CuffDiff is limited to A vs B, whereas DESeq2 and Sleuth are much more flexible.

Log in to answer this question.