Are TPM values generated by different programs comparable?

Question

[amy.bashir]

Hello!

I am currently analyzing a set of RNA-seq data. I first used StringTie to calculate the TPM values of reads that are mapped to the genome (a non-model organism). Then, for the ~5% of reads that were not mapped to the genome, I used Trinity to de novo assemble the reads and then used Kallisto to calculate the TPM of these transcripts.

Are the TPM values calculated by StringTie and Kallisto comparable? For example, can I combine the two lists of TPM values and look at the top 30 expressed transcripts?

Thank you very much!

Answer 1

[h.mon]

No, I don't think concatenating your TPMs like this is a good idea: both data sets are normalized to 1million, however the number of reads is very different between them.

If you want to do something like this, I would:

1) map reads to genome

2) used unmapped reads to assemble transcriptome

3) extract / get transcripts from genome

4) concatenate assembled transcriptome and extracted transcripts into new reference transcriptome

5) use kallisto to estimate TPMs / counts for new reference transcriptome

[amy.bashir]

Oh, that's a great idea. Thank you very much!