TPM normalization of transcript

Question

[Liftedkris]

Hi I’m new here and I’m happy to be here

I have a single end transcript and I want to normalize it to TPM. I have tried different ways but none of them has worked so far. I wanted to use Kallisto but kallisto requires me to know both the transcript length used in the sequencing and the standard deviation since my reads are single end. I don’t know the transcript length and the standard deviation since I didn’t carry out the sequencing myself and also the sequencers did not leave the information on GEO. Please what is the best tool to normalize this transcript to TPM

[ATpoint]

kallisto does not ask for transcript but for fragment length. That is the length of the cDNA fragment being sequenced. Typically, one would ask the wetlab staff who prepared the sample for this information. It is determined by Bioanalyzer/Gel/Tapestation. What is the dataset? I am pretty sure one can find the info there somewhere. Alternatively, look at salmon. It does not ask for this parameter in single-end mode.

[Liftedkris]

Thanks for your response, yes it is fragment length not transcript length. I’m not the sequencer and also don’t know who sequenced the data. Here are the data GSM 2779509, GSM2779516, GSM2779512 and GSM2779519. How do I know the fragment length and the standard deviation

[ATpoint]

Ok, the "method" section is very sparse and it does not seem to be associated with a publication. You can try the defaults, so 200 as length and 20 as SD or simply use salmon, which does not rely on that information. Also be careful with these kinds of data, as you literally have zero background information what this sample is.

Answer 1

[Kristoffer Vitting-Seerup]

Please note that Salmon DOES rely on fragment length and sd - it just have default parameters (which Kallisto used to have but they removed it to avoid mis-use like this).

If you dont know the fragment length you should probably not use tools that rely on that - use the old fashioned genome-mapping + quantification (STAR can fx do both simultaneously) and just calculate RPKM values instead.

In your case you cannot calculate TPM as TPM values are directly dependent on knowing the fragment length and sd. See this blog for a discussion of TPM (and other values)

[ATpoint]

My bad, you are right, Salmon has defaults that probably are approriate for most standard RNA-seq experiments.