MetaMP Ecosystem for Unified, Auditable, and Benchmark-Ready Data for Reliable Membrane Protein Annotation.

Experimentally resolved membrane-protein structures have increased substantially, yet annotations remain fragmented across resources differing in scope, curation criteria, and identifier conventions, complicating cross-database comparison and downstream analysis. We present MetaMP, a membrane-protein reconciliation and benchmarking platform that harmonizes metadata from MPstruc, RCSB PDB, OPM, and UniProt into a unified, searchable resource integrating 4,089 unique structures. MetaMP provides provenance-aware discrepancy analysis, quality-control workflows, and 2 assistive modules as reproducible baselines: (a) a broad structural-group classifier trained on OPM-derived membrane-orientation descriptors and (b) a transmembrane-segment benchmarking layer integrating sequence-based and structure-derived topology sources. Cross-source comparison identified 121 broad-group conflicts between MPstruc and OPM (2.96% of 4,089 harmonized entries). These contested cases were expert-reviewed to form a 121-record discrepancy benchmark. Under strict label matching, OPM agreed with expert annotations for 96 of 121 records (79.34%), while the MetaMP assistive classifier agreed for 25 of 121 (20.66%) and MPstruc for 17 of 121 (14.05%). Under benchmark-aware evaluation applying a biologically motivated label-collapsing rule, agreement reached 88.43% for MetaMP, 80.99% for OPM, and 78.51% for MPstruc. In a 24-participant task-oriented user study, structured training was associated with faster task completion, and the adapted SUS-style usability score averaged 72.81, placing the system in the above-average to good range. MetaMP is not intended to replace primary databases, but to make disagreement among them explicit, traceable, and biologically interpretable, providing a reproducible framework for annotation harmonization, expert-guided curation, and membrane-protein benchmarking. Source code and deployment materials are available at https://github.com/Ebenco36/MetaMP-Server.