TPM value difference of a sample between runs

[abhijitcbio]

Hi, Thanks for developing such a nice and easy to use tool.

I am currently analyzing a bunch of ONT long-read samples and trying to map them using isoquant.

I have two different cell lines Nalm6 and Jurkat with multiple conditions.

In the first run, I am processing only the Nalm6 samples and in the second run, I am running isoquant with both Nalm6 and Jurkat files together.

However, in the first run let's say for the "SON" gene I am getting the following TPM value for one of the Nalm6 condition

transcript_id	NALM6_EV
ENST00000356577	11.337801
ENST00000381679	1.223618
transcript19873.21.nnic	0.33119

But for the second run, I am getting the following values for the same sample -

transcript_id	NALM6_EV
ENST00000300278	0.971314
ENST00000356577	0.311926
ENST00000381679	1.101613
ENST00000436227	6.580062
transcript22541.21.nnic	0.311926

I am attaching the TPM table of SON transcripts across samples for the two runs. Expression_Difference.xlsx

I am using the .transcript_model_grouped_tpm.tsv file to get the values.

For your reference, I am attaching the log files from the two runs.

NALM6_JURKAT.isoquant.log NALM6_all_conditions.isoquant.log

Let me know if I am making any mistake in running the code.

Looking forward to your reply!

[andrewprzh]

Dear @abhijitcbio

I don't see any errors in your command lines. Here are a few comments:

• If you are looking for expression of reference transcripts, I'd suggest using *.transcript_tpm.tsv. transcript_model_tpm.tsv includes also novel transcripts.

• In general, TPMs may vary significantly across different runs, since reported transcripts and thus scaling factor may also differ. I think it makes sense to take a look at counts and grouped (per sample counts). Moreover, current version may not give very accurate results in transcript_model_tpm.tsv. So, it makes sense to check transcript_counts.tsv.

• Next release (3.4) will contain significant improvement is quantification.

Best Andrey

[andrewprzh]

IsoQuant 3.4 is released, so I suggest to give it a try.

[andrewprzh]

I'll close this issue for now, please, re-open if needed.