How to compare kallisto and stringtie output

Question

[yusuf]

Hi,

Can anyone help me in understanding how to compare kallisto and stringtie result on transcript level.

Kallisto result have some IDs like ENST00000004132 ENST00000004134.1 ENST00000004135.7 ENST00000004139.9

Kallisto result have some IDs like ENST00000004132 MSTRG9.41.3 ENST00000004135 ENST00000004139

So i am having trouble in comparing them. What is this xyz.1 or .7 as the given trancript id xyz only and in stringtie few new ids are MSTRG.9.41.3

it will be very helpful if anyone guide me how to compare it.

Thanks in advance. :)

[ATpoint]

Please use google and the search function for information about MSTRG. Plenty of posts on that e.g. here I think there is even a section in the stringtie manual about it (it is the newly assembled transcripts).

[yusuf]

Thank you. it was helpful.

Answer 1

[Devon Ryan]

You're comparing apples and oranges. The whole point of stringtie is to find novel transcripts (e.g., MSRTG.9.4.3), which will rarely if ever exist in human samples like yours. The various ENST ID's are from ENSEMBL and either contain revision number (the .1 or .7) or not. Revision numbers typically come with Gencode GTF files, I'm not sure how you managed to get some transcripts both with and without them.

[yusuf]

Thank you.

Should I cumulate transcript abundance into gene abundance and compare. as I need comparision between stringtie and kallisto.

Also new transcript but in later part of analysis. yes I got few in human data. luckily i say.

[Devon Ryan]

Most new human transcripts will be false positives. Comparing transcript-level quantifications is pretty common, since both tools are returning that (even if you end up using gene-level quantifications downstream).

Answer 2

[Kristoffer Vitting-Seerup]

If you want to compare Kalliso and StringTie you need to quantify the same set of transcripts. This can be done by either:

1. Run Kallisto on the transcripts produced by StringTie (from the resulting GTF after running StringTie -merge)
2. Run StringTie with "-e" (and "-B") option to avoid identifying new transcripts.

Cheers Kristoffer