Entering edit mode
I found a sample during QC that appears to be an outlier as it did not cluster with the other replicates in a multidimensional scaling plot. A pearson correlation also showed that it is more similar to the control. In limma I can use the voomWithQualityWeights method to specifically down-weight this sample amongst its replicates. I'm wondering if there a similar approach in DESeq2 instead of flat out removing it from the analysis?
Yes, you can supply weights (in a gene x sample matrix format, 0 to 1 valued), but we don't have a method like voomWithQualityWeights which estimates them for you.
So if I want to down-weight one sample across the experiment, would I supply a matrix that has all 1s except one column where it would be say, 0.5?
It's a gene-by-sample matrix so every gene per sample gets a different weight. The voom function you refer to combines the per-observation level weights with the sample weights (from arrayWeights), so it's not straight-forward to estimate/guess/makeup that manually to DESeq2. Would it be so bad to just use voom or remove the outlier and stick with DESeq2?