Hi all,

I’ve used tximport to import RSEM's “*.genes.results”  files and I’ve found that tximport uses the values from RSEM’s "expected counts" regardless of the value specified to the argument “countsFromAbundance” (i.e. “lengthScaledTPM”, “scaledTPM” or “no”). Would I have to correct the expected count values for the average transcript length per gene before I perform a differential gene expression analysis with limma-voom? 

Thanks, 

Celine

Thanks for noting this. If you use the latest version of tximport you can provide the isoform files to tximport(), and then you can specify countsFromAbundance of "lengthScaledTPM" or "scaledTPM". I guess I could build out more support to do this from the gene-level files specifically for RSEM, but I like having the clean interface that is mostly built around importing transcript-level data (and we now support RSEM).

I've added a warning to tximport() now if countsFromAbundance != "no" and txIn=FALSE, because it is just importing the data from the gene-level files, and not replacing the counts that RSEM computed.