Abstract
<title>Abstract</title> <p>Pocket-finding programs often identify several geometrically plausible candidates in oligomeric proteins, creating a practical prioritization problem for docking, simulation, and experimental follow-up. Interchain boundaries, nested subpockets, internal allosteric cavities, complex-dependent sites, and static-dynamic discordance further complicate this choice. We developed OIPS, a traceable framework that integrates heterogeneous evidence to prioritize candidate pockets. Five pocket-finding workflows yielded 1,742 records across a curated case series of 21 oligomeric protein-ligand systems. Reference-blind reconstruction excluded 178 records lacking both a center and residue set, consolidated 1,417 within-tool units, and generated 733 medoid-constrained cross-tool candidates. OIPS-Pstatic ranked candidates using prespecified modules for cross-tool consensus, geometry, ligandability, oligomeric relevance, and evidence completeness. Reference association, redocking, molecular dynamics (MD), and literature annotations were added only after static rankings were fixed. Automated quality control and rule-based annotation were applied to all 63 Top-3 candidates without expert adjudication. The first reference-associated candidate ranked first in 12/21 systems, within the Top-3 in 18/21, and within the Top-5 in 19/21 (mean reciprocal rank [MRR] = 0.723). The first candidate supported by reference association or assembly/dynamic evidence ranked first in 20/21 systems and within the Top-3 in all 21 systems (MRR = 0.976). Removing Orel increased reference Top-1 recovery from 0.571 to 0.667, while lowering reference Top-3 recovery from 0.857 to 0.810 and first-supported Top-1 recovery from 0.952 to 0.905. The ablation exposed a trade-off between retrospective ligand-location recovery and preservation of assembly-supported alternatives. Within this case series, OIPS provided an auditable basis for comparing competing pocket hypotheses, identifying evidence conflicts, and selecting targeted follow-up analyses.</p>