I have been using the DESeq2 package for differential expression and have really liked the flexibility that the package provides for complex models including interaction effects. In one of the recent updates some of the interaction term functionality that I had been using has been disabled. Specifically, the ability to test one interaction term versus all others within the expanded model matrix framework. I was wondering if there is a workaround to accomplish the comparison of one interaction term to all of the other interaction terms after this update.

My current workaround is to run the last version of DESeq2 that allowed testing of interaction effects within the expanded model matrix, which I think is 1.9.34, but I would obviously prefer to stay updated with the newest version of the software.

Using the current vignette example with conditions {A, B} and the genotypes {I, II, III}, I would like to fit the full interaction model (~ condition + genotype + condition:genotype) and test the condition specific effect of one genotype as compared to the rest of the genotypes. I believe that the correct way to do this is with this contrast: contrast=list("genotypeI.conditionB", c("genotypeII.conditionB", "genotypeIII.conditionB")).

I think that originally this was the correct usage of the interaction term comparison within the expanded model matrix. I understand that this has caused confusion for users, but I was wondering if it is possible to complete such a comparison under the current implementation, or if reverting to a previous version is the best way to go for this type of analysis. Thank for any help!

hi Marcus,

Sorry that we moved away from a design that was working for you. In order to make the code and documentation consistent and straightforward, we just went with only standard designs for interactions.

If you want to contrast one interaction term against all the others, it is still possible using standard model matrices.

Try a design of ~ genotype + genotype:condition

Then you will have a condition effect for each level of genotype, including the reference level.

You can constrast pairs of them using the list style of the 'contrast' argument.

If you want to compare one against the average of others, you can use listValues. e.g. listValues=c(1, -1/3) if you were comparing 1 level against 3 levels.