Dear Mike,

I was wondering about two scenarios of differential gene expression testing within DESeq2 framework. There are several components to consider -- condition (categorical; placebo or drug), survival time (continuous variable), other covariates that I want to control for (mostly categorial).

1. How would you implement design for the following question: expression of which gene is correlated with the survival time, specifically in the drug arm, compared to placebo?

2. Given (1.), how would you control for additional covariates, such as age, so that the question changes to: expression of which gene is correlated with the survival time, specifically in the drug arm, compared to placebo -- while controlling for the differences in age?

I'm specifically wondering about the practical implementation of the design formula within DESeqDataSetFromMatrix, and how would the contrasts used in 'results' function look like.

Thanks!

Do you have an idea how log(counts) ~ survival time? You will need to choose a parametrization, linear etc, or you could use the splines package in R as well. Looking at log normalized counts using plotCounts for example genes might be a useful way to start.