RNA-seq genes differentially expressed in combined treatment
2
0
Entering edit mode
@davidashbrook-10030
Last seen 8.7 years ago

Hi,

I have an experiment whereby Treatment A and Treatment B both cause a small difference in phenotype, but Treatment A + Treatment B together causes a large difference in phenotype, above what is seen by either alone. 

We have RNA-seq data from brain samples of saline treatment (control), treatment A, treatment B and treatment A+B, with 4-5 replicates each. I'm interested in seeing what differences in expression are due to the combination of A+B. 

I've used DESeq2 to do comparative analyses between each pairwise comparison,and used two approaches to try and get at the genes deferentially expressed in A+B. 

Firstly I looked at differential expression of A vs A+B, and then removed genes found to be significantly different in saline Vs B, to try and look for genes deferentially expressed only in A+B.
The second approach is to look at A vs A+B and B vs A+B and to investigate the genes found to be significant in both. 

I think these approaches works, but I may miss things (e.g. a small expression difference in A or B alone, but a much large difference in the combination).


Do the described methods make sense? Is there a better/more elegant way of doing this? 

I hope this makes sense, and thank you in advance,

 

DESeq2 deseq2 RNA-seq • 1.3k views
ADD COMMENT
1
Entering edit mode
@ryan-c-thompson-5618
Last seen 11 weeks ago
Icahn School of Medicine at Mount Sinai…

This sounds like the classic use case of an interaction model, which tests whether the combination of two or more treatments is different from the mere sum of the effects of each treatment on its own. This kind of model is described in the DESeq2 manual section 3.3, titled "Interactions".

ADD COMMENT
0
Entering edit mode
@davidashbrook-10030
Last seen 8.7 years ago

Hi Ryan,

Thank you for the reply.
I've had a look at the examples, and can't work out how to get my situation to fit. For example I've tried:

>dds <- DESeqDataSet(se, design = ~ TreatmentA + TreatmentB + TreatmentA:TreatmentB)

Which gives:

 

> resultsNames(dds)
[1] "Intercept"                  "TreatmentA_untreated_vs_treated"   "TreatmentB_untreated_vs_treated"  "TreatmentAuntreated.TreatmentBuntreated"

However, I'm not sure where to go from there, to get at what I want, i.e. The interaction between TreatmentA+Treatment B, compared to either A or B alone, without doing two comparisons and looking at the intersect. 

I'm not sure if it is possible, and looking at the intersect between TreatmentA vs A+B and TreatmentB vs A+B does seem to work, so maybe I'm just trying to make things too complicated. 

 

Thanks again,

 


 

 

ADD COMMENT
2
Entering edit mode

dear David,

If after reading the DESeq2 interactions documentation, you have further questions on the interpretation of coefficients in an interaction model, I'd recommend you discuss with a local statistician at your institute.

It's not so much a software question as a question of how these models work, and these are standard tools in statistics and linear modeling. Beyond writing up documentation and making figures as we have in the vignette section, the best thing to do is to sit in front of a whiteboard with someone who can sketch how these work.

ADD REPLY

Login before adding your answer.

Traffic: 525 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6