I have PRO-seq data with the following conditions: untreated (time 0), vehicle, and treatment (3 replicates each, 4 time points for each vehicle/treatment combination). I would like to evaluate treatment effects as a function of time. I have done the following:

1. Divided the reads in each condition by the corresponding DEseq size factor

2. Subtracted the vehicle mean from the treatment values at the corresponding time point (assuming additive effects)

3. Subtracted the min over the whole expression matrix (new min = 0)

I would like to use DESeq2 for DEG analysis, but clearly cannot enter the matrix of interest as counts. Would rounding be reasonable? Are alternative approaches advised?

Do not use subtracted values as input to DESeq2 or any count method.

You can account for baseline values using interaction models, and keep the data on the original count scale. See the interaction section of the DESeq2 vignette.