Question

Transcript-level differential expression using DESeq2

1

Entering edit mode

Rain Yi ▴ 10

@rain-yi-24046

Last seen 15 months ago

Dear DESeq2 community,

I am just wondering whether I can use DESeq2 to perform transcript-level differential expression with Salmon quantification data? And if I can, is that essential to include bootstrap when using Salmon? I saw a tutorial said DESeq2 will not work if one want to do transcript-level differential expression, just curious whether it is true or not.

Thanks.

deseq2 transcript-level differential expression • 860 views

ADD COMMENT • link updated 3 days ago by Michael Love 36k • written 21 months ago by Rain Yi ▴ 10

score 1 · Answer 1 · 2020-08-25

1

Entering edit mode

Michael Love 36k

@mikelove

Last seen 27 minutes ago

United States

We have two relevant papers from my lab here:

This is a benchmark that explores gene level tools used to perform DTE (the benchmark follows the workflow section on DTU):

https://doi.org/10.12688/f1000research.15398.3

And this is a new Bioconductor method/package we developed (based on ideas from SAMseq) for performing DTE. The method is called Swish and the package is fishpond.

https://doi.org/10.1093/nar/gkz622

I really like this method, it’s fast, easy to use, has very good error control and sensitivity across a range of sample size, and the vignette shows how to make results tables and plots just like in DESeq2 (counts plots and MA plots).

https://bioconductor.org/packages/release/bioc/vignettes/fishpond/inst/doc/swish.html

ADD COMMENT • link 21 months ago Michael Love 36k

0

Entering edit mode

Thanks Michael. So what you suggested is using fishpond to do transcript-level differential expression instead of DESeq2?

ADD REPLY • link 21 months ago Rain Yi ▴ 10

1

Entering edit mode

Yes, it is exactly designed for this. Give it a shot and let me know if you have any questions.

ADD REPLY • link 21 months ago Michael Love 36k

0

Entering edit mode

Thanks Michael! I'll try it out. One more question, can I use DESeq2 for transcript-level differential expression if the read quantification data were generated by StringTie since this is an alignment-based aligner? Or I still have to use fishpond?

ADD REPLY • link 21 months ago Rain Yi ▴ 10

1

Entering edit mode

With StringTie they don’t have a measure of quantification uncertainty so you can’t use Swish. You can use DESeq2 then.

ADD REPLY • link 21 months ago Michael Love 36k

0

Entering edit mode

Hi Michael!

1)I have done quantification using Kallisto. Can fishpond be used with txi.kallisto$counts as an input to perform transcript-level differential expression.

2)In the vignette, https://bioconductor.org/packages/devel/bioc/vignettes/tximport/inst/doc/tximport.html#kallisto, abundance.h5 is used for reading transcript-level information. If I do not use tx2gene and follow the DEseq2 protocol, will I get transcript-level differential expression?

thanks ekta

ADD REPLY • link 8 days ago Ekta • 0

0

Entering edit mode

You can use tximeta to import kallisto bootstraps to run Swish as well. You will get transcript level counts for performing DTE.

ADD REPLY • link 8 days ago Michael Love 36k

0

Entering edit mode

Hi Michael .. I used kallisto bootstraps to run Swish and performed the Differential transcript expression.

sum(mcols(y)$qvalue < .05) gives 1317

But surprisingly the qvalue for all these 1317 differentially expressed transcripts is 7.59E-06.

And after this, 220 transcripts have qvalue 0.916510084, 39337 transcripts have qvalue of 0.9442869.

What could be the possible reasons for it?

Ekta

ADD REPLY • link 6 days ago Ekta • 0

0

Entering edit mode

This is not an issue, see the vignette which talks about this phenomenon.

ADD REPLY • link 3 days ago Michael Love 36k