Peaks at every promoter with Input material

[ETHANol]

Looking at my ChIP-seq data, the peaks on the INPUT samples are all on promoters. What would cause this? Is a well prepared input sample usually so uneven. I would have expected it to be more flat. This obviously makes the peak calling for the ChIP material difficult.

Is it likely the sonication or the data analysis?

[kopi-o]

You might want to look at this paper

http://www.plosone.org/article/info:doi%2F10.1371%2Fjournal.pone.0005241

[ETHANol]

Thanks!!!

That pretty much explains it. I just never would have guessed there was such a strong preference for generating fragments at TSS.

[ETHANol]

I guess the question is what, if anything, can be done to minimize bias in the input sample?

[Chipper]

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1891346/

FAIRE-seq is like ChIP-input bút without protK digestion, and profiles are similar to what is in the plosone paper. Another possible cause would be size selection if open chromatin more easily is sheared into short fragments, but I am not sure if this has been shown?

[captainentropy]

@chipper. I do believe that's the case. A neighboring lab where I am is interested in these things and from what I understand it looks as if the shorter fragments map to euchromatic regions and the larger to more heterochromatic regions (which is why I don't like the Illumina library size selection of about 175-225bp. I go 200-600bp). So this makes sense actually because euchromatin is less packed than heterochromatin therefore you would expect to find fewer protein-protein crosslinks. This would favor shearing in the less crosslinked regions. It would be the exact opposite with the heterochromatic or otherwise densely packed regions.

It should be nothed though that it's also possible the phenomenon is a result of incomplete decrosslinking. DNA that is associated with proteins that are lightly crosslinked would more likely be liberated in the decrosslinking step. AFAIK no one has really tested the decrosslinking test. Regardless, it would give the same result. It's also and important thing to keep in mind for the crosslinking step - don't overcrosslink your sample. It's quite easy to do actually.