Null hypothesis experiment to evaluate SINATRA Pro’s ability to find regions of interest (ROI) in each of the proteins analyzed in this study.

Here, we assess how likely it is that SINATRA Pro finds the region of interest (ROI) by chance. These ROIs include: (i) the Ω-loop (residues 163–178) in TEM; (ii) the flap region (residues 47–55) in HIV-1 protease; (iii) Domain 2 (residues 208–308) in EF-Tu; and (iv) the DFG motif (residues 381–383) in Abl1. Note that protein structures were only analyzed if they contained an entire ROI. For example, in the context of Importin-β, the superhelix includes the entire structure and so we do conduct a null analysis. In this experiment, to produce the results above, we generate “null” regions on each protein using a K-nearest neighbors (KNN) algorithm on different atoms as random seeds [84], and exclude any generated regions that overlap with the ROI. Next, for each region, we sum the association metrics of all its atoms. We compare how many times the aggregate scores for the ROI are higher than those for the null regions. These “P-values,” and their corresponding calibrated Bayes factors (BF) when the computed P < 1/e, are provided above. Note that P-values less than the nominal size 0.05 and BFs greater than 2.456 are in bold. Results above are based on SINATRA Pro using parameters {c = 20, d = 8, θ = 0.80, l = 120} while varying the radius cutoff parameter r for mesh construction on each protein structure.