I am doing a differential gene expression (DGE) analysis using DESeq2 v1.24.0 with ~200 bulk RNA-seq samples (whole-blood) with a continuous variable of interest. I am interested in applying count-based outlier detection and replacement using Cook's distance. For factors, this functionality is applied by default when calling DESeq(). However, based on the DESeq2 vignette, outlier detection and replacement is disabled for continuous variables:

Note that with continuous variables in the design, outlier detection and replacement is not automatically performed, as our current methods involve a robust estimation of within-group variance which does not extend easily to continuous covariates

Are there plans to include outlier detection and replacement for continuous variables in a future release of DESeq2? Has anyone else encountered this issue and constructed a good workaround to apply Cook's distance and trimmed mean replacement while still being able to re-run model fitting using DESeq2 functions?

I don't have plans to add anything now to deal with outlier replacement for individual observations in designs with continuous variables. I think you can certainly detect them with the Cook's distance matrix we provide, but I don't feel comfortable replacing them with null values / don't have code for that.

edgeR has a method for down-weighting individual observations from 2014, which you might check out:

https://academic.oup.com/nar/article/42/11/e91/1427925