For each TF, create one positive sequence set with all the datasets of this TF in this experiment (e.g. CHS, GHTS, ...), and one negative sequence set with all the datasets of the other TFs in the same experiment.
This will mimic the "aliens" benchmarking in IBIS
'aliens': peaks of non-related proteins not overlapping any reproducible peaks of the target transcription factor;
Remarks
Due to time limitations, we will not filter out the overlapping peaks.
We generalize the idea by applying it also to non-genomic datasets
For PBM, we use a different approach by defining "others" as the background spots , i.e. the 3500 spots having the lowest binding signal
For each TF, create one positive sequence set with all the datasets of this TF in this experiment (e.g. CHS, GHTS, ...), and one negative sequence set with all the datasets of the other TFs in the same experiment.
This will mimic the "aliens" benchmarking in IBIS
Remarks