A Rigid-Corridor, Water-Relay Model of Allosteric Communication in Ubiquitin from Short Tree-Search Dynamics

Understanding how allosteric signals propagate through a protein without passing through the active site remains a central challenge in structural biology. Here, we apply a hybrid Monte Carlo Tree Search (MCTS)–Langevin dynamics algorithm to ubiquitin (PDB: 1UBQ), a 76-residue β-grasp protein whose conformational plasticity underlies its roles in proteasomal targeting, DNA repair, and cell-cycle regulation. We ran 500 MCTS iterations with 200 Langevin steps per rollout using the AMBER ff14SB force field with GBN2 implicit solvent, producing 51 frames spanning the conformational landscape near the native state. Four complementary analyses were applied to the resulting trajectories. Principal component analysis (PCA) of all 660 heavy-atom positions identified a dominant conformational mode (PC1, 65.0% of variance) centred on the GLU18–ASP21 loop and ASP39, with 90% of all variance captured in only five PCs. Graph-theoretic betweenness centrality of the ensemble-averaged contact network revealed Thr12 as the single most critical allosteric bridge residue (BC = 0.156), followed by LYS33 and ARG54, both of which appear in the five shortest allosteric paths computed by Yen’s algorithm. Mutual information (MI) analysis of pairwise Cα–Cα distance computed with a k-nearest-neighbor estimator and bootstrap uncertainty quantification uncovered correlated motions between residues 24–32 and the C-terminal tail (residues 74–75) that are not predictable from contact frequency alone (Pearson r = 0.11). Hydrogenbond analysis using proper donor–acceptor criteria showed that HOH134 maintains a persistent interaction with LEU56. All five computed allosteric paths converge on HOH117 and HOH134. Cross-PDB analysis of 12 high-resolution ubiquitin structures confirms these water positions are conserved (occupancy > 0.8 in 10 structures), consistent with their acting as structurally stable relay stations. Finally, stability analysis of all 59 crystallographic water molecules showed that HOH117 and HOH134, which terminate every computed allosteric path via LEU56→HOH117→HOH134, are significantly more stable (RMSF = 1.26 and 1.34 Å, respectively) than the protein heavy-atom mean (1.81 Å). Together, these results generate the testable hypothesis that allosteric communication in ubiquitin flows along a small number of rigid, topologically central residue corridors and is relayed to the solvent through a pair of structurally conserved buried waters, providing tractable targets for future perturbation and inhibitor design studies.