I'm currently working on ATAC-Seq analysis and I have to do Differential Expression analysis. I'm pretty new to ATAC-Seq analysis so I'm sorry if my question is dumb.

I did read through the DiffBind vignette and I see that Deseq2 was said to be more preferred. When I tried running my data using both EdgeR and Deseq2, I saw a difference of about ~2000 peaks. There was also multiple possible ways of normalizing the data each of which gave a different set of results. I also saw a post which said you should not normalize the data and input it to edgeR since it takes raw read counts, but whereas they advise to do so in the vignette. So, I'm a bit confused with the following:

1. Should I normalize the data before inputting it into edgeR or Deseq2.
2. If so, which type of normalization should I use? [Our data does not have spike-ins, so I was thinking of resorting to following as per the vignette: background=TRUE and normalize=DBA_NORM_NATIVE]