Hi,

I have a dataset of ribo-depleted RNASeq from 3 pairs of cell populations (with replicates). We have reasons to believe there will be abnormal expression at specific genomic regions in some of those populations which may not be consistent with underlying biological structure (e.g. intergenic expression).

In order to systematically look for expression in those regions, I have quantified the read density using Bedtools coverage in discrete 1kb windows to be able to directly compare the same positions between the cell populations (5000 windows per region).
I am trying to identify which of these 1kb windows is significantly differentially expressed. Would it be appropriate to use DESeq2 for this purpose, eg using DESeqDataSetFromMatrix? So the count matrix will be these fixed windows rather than variable length genes. Or is this incompatible with the model DESeq2 is based on?

Many thanks!

I don't see a problem with this (see also the csaw method and paper).

The most important consideration here is the size factors. I recommend looking at MA plots to ensure that the scaling is reasonable. That is, finding which regions did not change across condition.