Collapsing Transcript Count Matrix to Gene Count Matrix

Question

[marionette.kent]

Hi all,

My colleague is trying to do some DGE. He got hold of a transcript count matrix from another colleague but he was trying to do the analysis at gene level not transcript level. Eventually I found those RSEM files on a data portal. But I am wondering, in case I didn't find them, if it is safe to just add the counts of all transcripts from the same gene to get the gene level count, and what information is needed other than a transcript-to-gene table e.g. transcript length. Because the numbers don't add up to a whole number.

Many thanks

[Ram]

Please explore the tximport package. It should have a way to read in RSEM transcript level data and output gene level metrics.

[marionette.kent]

Hi Ram, do you know if that works even with an already compiled transcript count table? like a tsv file with columns of samples and rows of transcripts? I know it works if I have file outputs rsem.isoform.results.gz. But I will have a look at it again.

[Ram]

I don't think it will work with a count matrix, as all relevant info required to rebuild gene level metrics is lost.