Add score per million normalization when merging peaks

[GilbertHan1011]

In some cases, read depth or quality in cell types / samples may vary, which affect peak calling. Peak normalization introduced in this paper may help

[kaizhang]

I'm aware of this approach. But I'm not particularly interested in merging this feature now as I do not think this is the best way for solving this problem.

[adamklie]

Hi Kai and Gilbert,

Probably not the best place for this and happy to move it to a discussion, but thought I'd chime in since I'm dealing with this problem with a dataset. Have you considered giving an option for downsampling fragments to some uniform or maximum depth prior to peak calling? Is there any easy way to accomplish this with the current version of the tool?

Kai, I'm also curious if you have a best way of dealing with this in mind.