Hi all,

I have generated factors of unwanted variation (W_1) via RUVseq/RUVg (with spike-ins), but am having trouble incorporating them into ImpulseDE2. The reason I would like to use ImpulseDE2 because the literature suggests it is a good model for impulse-like treatment conditions and many time points, which our experiment had: we treated trees with a hormone once at the beginning of the timecourse and harvested them every day for 16 days.

In DESeq2, incorporating W_1 would be fairly straightforward, as I would run a LRT with a full model: ~W_1 + time + treatment + treatment:time and a reduced: ~W_1 + time + treatment. However, I am not sure how to incorporate W_1 appropriately into ImpulseDE2. Two options would be to add them as size factors or confounding variables, but I'm not convinced either are appropriate, as my understanding is that W_1 accounts for more than just library size variation and DESeq2 treats W_1 as a covariate, not a confounder. However, I'm not sure if there is a difference here, or if I am misunderstanding.

Any suggestions or clarity would be much appreciated! Thanks for reading.

In case anyone is interested or has this issue, incorporating W_1 into the vecConfounders argument effectively works as a "+" in the model, as verified by the developer.