Background regions for de novo Motif Discovery?

Question

[ltosti88]

Hello everyone,

I'm trying to identify the binding motif of a transcription factor based on ChIP-seq peaks identified with MACS2.

I tried different tools (among which Homer, MEME-ChIP, STEME, RSAT) but I have a couple of questions as I'm not sure about the validity/reproducibility of my results:

• Do you select the peaks (for example taking the top 10-20% of the peaks ranked by fold change/p-value) or you give as input the full list of them? What's the best practice?

• Do you use your own background sequences or you let the software generate its own background?

• In case you use a custom background, what do you use? For example, I tried to give randomly picked regions identified by DNase-seq, but this completely abolished the enrichment of the motifs that I get when I don't give any custom background.

Thanks for your help!