DESeq2 with 2x2 crossover design
0
Entering edit mode
vincev454 • 0
@fc21398a
Last seen 21 days ago

Hi everybody,

I have a crossover study 2x2 design with 2 treatments and 2 baselines. Each phase consisted of one measurement before treatment (pre or baseline) and one after treatment (post). The data table looks like this:

Subject | Sequence | Phase | Treatment | Time
   1         AB        1         A        pre-baseline     
   1         AB        1         A        post     
   1         AB        2         B        pre-baseline      
   1         AB        2         B        post     
   2         BA        1         B        pre-baseline
   2         BA        1         B        post
...

I'm really not sure about the formula i should use to compare the effect of treatments at final time point (post), controlling for baselines and also for the "Sequence" variable (as my PCA shows distinct separation based on this variable). I went through the vignette but didn't find any example with such a design, does somebody have an idea?

Any help would be greatly appreciated!

DESeq2 crossover • 120 views
ADD COMMENTlink
0
Entering edit mode
@mikelove
Last seen 1 hour ago
United States

You would do it the same as you would approach it with lm(), there are lots of posts online if you search for how to use linear models to approach this. Otherwise, I'd recommend consulting with a statistician, unfortunately I'm not in a position where I have sufficient time to work out designs, but have to reserve my time for software related support here.

ADD COMMENTlink
0
Entering edit mode

Thanks for you reply Michael!

So using lm(), i would use the following formula:

phyloseq_to_deseq2(phyloseq_object, ~ Treatment + Time + Sequence + Treatment:Time) 

Then, if i'd like to compare treatments at the final time point, i guess I would run something like:

results(diagdds, name="TreatmentB.Timepost", test="Wald", pAdjustMethod = "fdr")

But does "TreatmentB.Timepost" controls for baselines and "Sequence" variable in this case?

ADD REPLYlink
0
Entering edit mode

Selecting the appropriate statistical design is really on the user, I don't have sufficient time for consult and interpretation on the support site. That is the design I would use for a 2x2 crossover.

ADD REPLYlink
0
Entering edit mode

Thanks Michael!

I understand you can't advise users on statistical design, but could you tell me if, when using the design described above, the DESeq2 result "TreatmentB.Timepost" will take into account baseline values and "Sequence" covariate?

ADD REPLYlink
0
Entering edit mode

I’m really busy these days and don’t have time to advise on statistical designs. It’s really a good idea to work with a statistician who can help intercept meaning of coefficients in your design.

ADD REPLYlink

Login before adding your answer.

Similar Posts
Loading Similar Posts
Help About
FAQ
Access RSS
API
Stats

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

Powered by the version 2.3.3