How to normalize cut-site probability between samples

[kshimizu64928]

Hello,

I want to compare cut-site probability between samples.
However, I found that the backgrounds of cut-site probabilities of each sample were linearly correlated with mapped read counts of each sample.
Background was calculated as mean cut-site probability sufficiently far from motif (+5kb ~ +10kb).
Do you have any ideas to normalize cut-site probability between samples?

Thanks in advance,