Paired t-test with DESeq2
3
0
Entering edit mode
Mozart ▴ 20
@mozart-20625
Last seen 22 months ago

Hello there, I know you are not recommending to perform t-test when doing rna-seq but I really need to know how to perform the latter because the core facility who analysed my samples did this and apparently don't want to disclose the code used at this stage. The reason why I am so keen is that, when validating a few genes via qPCR here there is indeed a significant change for the genes obtained with this approach (t-test). Those genes didn't show up with the normal approach.

I guess they are using tximport as scaledTPM and then counts are used outside, R in excel to perform a t-test? Is that possible?

regards,

deseq2 • 365 views
1
Entering edit mode
ATpoint ★ 1.4k
@atpoint-13662
Last seen 4 hours ago
Germany

Others might give you a statistically-sound response but if you do not have access to the code or any details beyond doing a t-test on your data I would not use these results at all. Imagine a reviewer asks you at some point to clarify about the analysis strategy, what would you respond? Maybe you would need to re-do the analysis during revision which is the last thing you want. Don't blindly trust results. Either insist on the code and check if it is in agreement with current best practices or (preferred IMHO) get the raw data and perform a standardized analysis as suggested in the vignettes of the established tools (edgeR/DESeq2/limma). Unless you have a tricky design or a non-standard RNA-seq setup that requires custom adjustments you can pretty much follow the code suggested in the manuals. Poor agreement between qPCR and a potentially flawed analysis is probably not too unexpected.

0
Entering edit mode

Thanks ATpoint; I realised I made a mistake in the writing of this post. I meant to say that there's indeed a correlation between genes found with this approach (t-test) and qpcr. That's the rationale behind this post.

thanks.

0
Entering edit mode
@mikelove
Last seen 3 days ago
United States

0
Entering edit mode

Thanks for your reply. I am wondering how one could perform t-test with DESeq2. Don't know the code, unfortunately.

0
Entering edit mode

It’s not possible.

Please take a look at the abundance of documentation and the 2014 paper.

0
Entering edit mode
swbarnes2 ★ 1.1k
@swbarnes2-14086
Last seen 7 hours ago
San Diego

I really need to know how to perform the latter because the core facility who analysed my samples did this and apparently don't want to disclose the code used at this stage.

This isn't going to end well.

If you are responsible for this project, you do the analysis, so when people have questions, you can tell them what you did.

Maybe the core group just did something simple they knew how to do to give you something, rather than doing things the right way.