Huge TPM values in some tissues of an expression atlas

Question

[lessismore]

Dear all, I'm working with expression dataset with TPM values for each gene in several tissues. I've found that some genes range from 20 TPM in some tissues to crazy values such as 160.000 TPM in other tissues (e.g. in at least 5 tissues this gene is > 100.000 TPM). Importantly, this happens in all biological replicates of each tissue, so it's reproducible. Do you think this is reasonable? Thanks for the support

Answer 1

[jared.andrews07]

This is likely due to a mix of composition bias and very highly expressed transcripts not being all that sensitive to changes in levels. Often, these are house keeping genes or non-coding transcripts that are pretty highly expressed in pretty much all tissues.

For example, here's how GAPDH looks in GTEX:

Pretty high across the board, though with a fair amount of variation. MALAT1 is another good example. There are just a lot of reads that go to these transcripts, and reads are a finite resource, so counts for other genes are impacted (and these differences in composition are modeled during differential expression analyses).