transcription expression matrix for DESEQ2

Question

[Duckula]

Dear all, I am kind of new to DESEQ2 and I have an expression matrix csv that is consisted of ensemble transcription IDs (not gene ids) and their expression value in each condition. I am wondering if I can use this matrix directly as an input in the deseq2 pipeline.

Thanks Cheers

[andrew.j.skelton73]

Where did this expression matrix come from? - You can provide matrix input for DESeq2, but this must be integer counts of genes. For transcript level count estimates, you need to follow the tximport protocol, from either Salmon or Kallisto.

[Duckula]

Hi, Thanks for your reply. These are actually raw transcript counts that are coming from bitSEQ. The values are integer. so would be possible using tximport package?

Answer 1

[andrew.j.skelton73]

bitSeq is more an all in one for differential transcript expression. You wouldn't be able to input those estimates into DESeq2, I'd recommend following one of these methodologies if you want to do differential gene expression or differential transcript expression from the same quantifications:

Salmon

• Quantify your raw fastq with Salmon
• Read in using tximport
• Differential gene expression with DESeq2 or Limma
• Convert Salmon output for compatibility with Wasabi
• Run Sleuth for Differential Transcript Expression

Kallisto

• Quantify your raw fastq with Kallisto
• Read in using tximport
• Differential gene expression with DESeq2 or Limma
• Run Sleuth for Differential Transcript Expression

[Duckula]

Dear Andrew, Thanks a lot for your reply. The thing is that in the paper that this dataset is originally published they have mentioned they have put the raw transcription data into deseq for further analysis. That is why I was wondering how can I do it and how should I proceed with this matrix. Do you have any idea that might be worth to try ?

[andrew.j.skelton73]

In this case, it's probably worth contacting the author, or looking at potentially more detailed methods in the supplementary. Keep in mind, "just because you can, doesn't mean you should" - It's worth critically assessing what DESeq2 requires as inputs, against what they've provided from bitSeq.