Interaction Persistence-Based Identification of Key Binding Residues in the Cellular Retinol-Binding Protein 1 Complex

Identifying residues that form dynamically persistent interactions with a ligand is a central challenge in computational protein–ligand analysis. Although molecular dynamics (MD) simulations capture continuous conformational trajectories, existing analyses based on contact frequency, energy decomposition, or static distance criteria do not directly quantify the temporal stability of individual residue-ligand interactions. Here, we present a dynamics-based computational framework that quantifies interaction memory persistence in protein–ligand complexes using two normalized interaction descriptors, a distance-based proximity metric and a dipole-based alignment metric, each evaluated as continuous time-dependent variables along MD trajectories. The autocorrelation function (ACF) of each metric was computed for all residues, and the resulting decay profiles were decomposed into fast and slow relaxation modes via bi- or single-exponential fitting. A three-stage hierarchical filtering protocol integrating interaction strength, exponential fit quality, and slow-mode fractional amplitude was applied to identify residues exhibiting dynamically persistent interactions. The framework was applied to a 200 ns equilibrium MD trajectory of the Q108R CRBP1- all-trans retinoic acid (atREA) complex. The slow relaxation times (τ slow ) extracted from the proximity and alignment metric reached a mean of 14.5 and 10.3 ns, approximately 1 to 2 orders of magnitude longer than typical internal protein motions, reflecting binding-relevant interaction persistence. Thirty residues were identified in the proximity-based group and 19 in the alignment-based group, with 8 structural-dynamical hotspots, including I77, W106, and R108, recovered at their intersection. Scatter-map analysis showed that I77 and R108, reported as key binding-site residues, occupied prominent positions in the interaction-dynamics parameter space, consistent with their roles as a hydrophobic clamp and a chemical anchor, respectively. This modular pipeline is directly applicable to diverse protein–ligand systems and offers a quantitative basis for residence time-oriented drug design.