Title: Prediction of allosteric sites and mediating interactions through bond-to-bond propensities

Abstract: Allosteric regulation is central to many biochemical processes. Allosteric sites provide a target to fine-tune protein activity, yet we lack computational methods to predict them. Here, we present an efficient graph-theoretical approach for identifying allosteric sites and the mediating interactions that connect them to the active site. Using an atomistic graph with edges weighted by covalent and non-covalent bond energies, we obtain a bond-to-bond propensity that quantifies the effect of instantaneous bond fluctuations propagating through the protein. We use this propensity to detect the sites and communication pathways most strongly linked to the active site, assessing their significance through quantile regression and comparison against a reference set of 100 generic proteins. We exemplify our method in detail with three well-studied allosteric proteins: caspase-1, CheY, and h-Ras, correctly predicting the location of the allosteric site and identifying key allosteric interactions. Consistent prediction of allosteric sites is then attained in a further set of 17 proteins known to exhibit allostery. Because our propensity measure runs in almost linear time, it offers a scalable approach to high-throughput searches for candidate allosteric sites.