Abstract
<jats:p>The analysis of protein tunnels is essential for elucidating the dynamic variations of substrate entry, product release, and molecular transport pathways. In molecular dynamics simulations, however, continuous conformational motion shifts tunnel coordinates, reshapes bottlenecks, and redistributes pathway populations. Across heterogeneous molecular dynamics ensembles—independent trajectories spanning different conformational states, sampling protocols, or related protein variants— the identity of tunnels is obscured by fragmentation across frames and spatial clusters, making cross-ensemble tunnel correspondence and mutation-induced remodeling difficult to resolve. We introduce \textit{TopoTunnel}, a topology-guided framework that establishes protein tunnel correspondence across heterogeneous molecular dynamics ensembles by representing paths through local structural context rather than raw coordinates. Each tunnel point is encoded by a surface-gap-ranked four-atom descriptor, and complete paths are projected into a shared reference conformation. Constraint-based matching then retrieves corresponding path ensembles based on residue composition, exit position, and path length. To facilitate interpretation, coordinated visual analyses connect these pathway dynamics to bottleneck residues and network organization. Experimental results show that \textit{TopoTunnel} is invariant to rigid-body transformations, retrieves cross-state path ensembles with greater source consistency than the dominant-overlap CAVER cluster comparator, and reveals mutation-associated residue-network rewiring that is obscured by single-residue statistics.</jats:p>