When Does Higher-Order Representation Improve Protein-Complex Recovery? A Leakage-Controlled, Matched-k Comparison of Graph and Penalized Hypergraph Spectral Clustering.

This study presents a systematic comparison of graph- and hypergraph-based spectral clustering methods for protein complex identification across nine human protein interaction datasets. Using leakage-controlled train–validation–test splits, matched cluster-number evaluation, 25 paired random seeds, structural null models, and Gene Ontology enrichment analyses, the study examines whether preserving higher-order protein associations improves biological complex recovery. The results show that penalised hypergraph clustering provides significant improvements in several datasets, although the advantage is dataset-dependent rather than universal. The work also evaluates functional coherence across Biological Process, Molecular Function, and Cellular Component ontologies, offering a reproducible framework for assessing higher-order network representations in computational systems biology.