Still somewhat new to handling transcriptomic data, and have a newbie question.
I'm just trying to convert some RNA-Seq count data to TPM for the purpose of presenting qualitative comparisons about relative expression of various genes in a single cell type/condition. I'm aware that for comparisons between conditions TPM is not optimal.
I have a matrix of both raw counts for each gene/condition, and corrected counts generated by DESeq2.
My question is, could I use the DESeq2 corrected count data for the TPM conversion? My reason for doing this is that I have heard that DESeq2's normalisation method is more robust when correcting for library size and RNA composition than the TPM method. Obviously since TPM normalises after adjusting for gene length, it will still change the relative quantities slightly, but from a slightly better starting point?
Just wanted to check I'm not missing some obvious reason why this would be worse than just doing the TPM conversion on the raw count matrix?
The problem with TPM conversion is that many genes have multiple transcript isoforms, and the expression of these isoforms can change between conditions. To account for this workflows such as Salmon + tximport will quantify at the transcript level and then generate an aggregate gene TPM per gene using this transcript level information.
Cheers, this is good to know for future reference! Currently however I don't think that will work for me as I don't have the underlying sequencing data, just gene-wise count values.