Fig. 3
From: Leveraging protein quaternary structure to identify oncogenic driver mutations
Statistic construction. Here we consider up to three spheres (s∈{1,2,3}) and radii of either 3 or 9 Å. The first step is to calculate μ and σ over each column and then a normalized statistic \(Z_{{i,s,r}} = \frac {X_{i,s,r }- \mu _{s,r}}{\sigma _{s,r}}\) for each cell. Then the maximum is taken over each row, specifically m a x s,r Z i,s,r , to obtain Z 0,…,Z 1000. One minus the percentage of cases where Z 0≥Z i , for i∈{1,…,1000}, is the p-value of our observed statistic Z 0. As 1000 simulations were run, if Z 0>Z i ∀i, a p-value <1.00E−03 is reported [26]