The state space of GC content and the log of the average accessibility of peaks is transformed by the Mahalanobis transformation in order to remove the correlation between the two variables. This transformed space is split into an even grid of bins with a specified number of divisions (50) along each axis evenly spaced between the minimum and maximum values. For a peak in a given bin j, the probability of selecting another peak x in bin i is given by: P(x∣x∈bi)=f(d(i-j)∣0,w)ρi
