Is it possible to run diffbind with my "own" binding affinity table? I might want to delete certain peaks or have my own way of combining peaks across replicates.

Let's say I have a table mytable.tsv that looks like this:

       sample1    sample2    sample3
peak1      312        366         12
peak2      999         12        512
peak3        2        521         52

If so, should I then still do e.g. TMM normalization afterward? Or should I load a TMM (or whatever normalization I want) normalized table?

You can either load your own set of peaks and have DiffBind do the counting, or load your own set of counts. In either case, you should use raw read counts and the data will be normalized when dba.analyze() is invoked.

To use your own set of consensus peaks, simply use the 'peaks=' parameter in 'dba.count()`.

Alternatively, to load raw counts, you can include a Counts column in your sample sheet, which will be the path of a file containing the counts for each sample. Possible formats for these files are documented on the help page for dba.peakset(), for the counts parameter.