Comparing input peaks to IP peaks

[biznatch]

I have some peaks in my ChIP data where the height is eg. 100, and the input at the same site is also around 90-100. These sites tend to be at regions with low mapability and/or repeat sequences, as shown with the mapability and repeat masker tracks in the UCSC browser. I assume these peaks are background and don't consider them real.

I have other sites which will also have a height of 100 and the input will have what looks like a small peak, maybe 10-15, while the input for the rest of the region is <5. Sometimes the IP peak is really big, like 200-300 and the input is maybe 20. These sites are generally in regions with no problems such as repetitive sequences.

Can I consider this second group of peaks, the ones that have a small input peak, as real? Maybe they're real but just not quite as high as they appear in the IP sample, ie. some of the enrichment in the region is background but most is real?

[ETHANol]

It's all about the ratio of the input to ChIP. You will tend to have 'peaks' in your input in areas of open chromatin as long as the peaks in your ChIPed sample are higher then you have enrichment. Some techniques are more biased in terms of uneven fragmentation then others. The more even the fragmentation the cleaner the analysis is.

But yes you can consider those peaks. Most peak finding programs take this into account.

[biznatch]

Ok thanks that is good to know. When I've used peak finding programs which take into account the input sample they do count these as peaks in the ChIP sample, I just wanted to make sure that from a biological (as opposed to statistical I guess) perspective it makes sense.