Hello,

I'd like to preface this question by expressing my appreciation for this tool and the hard work that has gone into creating it. The flexibility in hypothesis testing that it provides is already amazing. So, thank you.

What I'd like to know is if it is possible to test a joint hypothesis within DESeq2?

For a use case, consider my experimental setup:

I am looking at a parameter which can be "low" or "high" at two timepoints "early" and "late". I am interested in, for example, the genes that show significant LFC enrichment for being in the "high" group at timepoint "early" AND significant LFC enrichment for being in the "high" group at timepoint "late".

Currently, I am just finding the "significant" (i.e., below some arbitrary adjusted p-value cutoff) contrasts for the different timepoints separately and looking at which genes are in both lists. But, given the power of the modeling provided by DESeq2, it seems like it would be a better idea to somehow encode this hypothesis up front and get a single p-value.

Thank you very much, Adam

Thanks for the note. Appreciated!

We don't have a combination test or joint test for requiring two coefficients to be non-null. I would just combine the two FDR sets by taking the intersection, as you say.