ECHO-PPI: Trustworthy AI makes protein module detection interpretable

ECHO-PPI, developed by Sima Soltani, Mehrdad Jalali, and Yahya Forghani, addresses a key limitation of existing community-detection methods in protein-protein interaction networks: the lack of interpretability. While current methods can recover candidate protein complexes, they rarely explain why a protein is assigned to a specific module or whether that assignment should be treated as core, peripheral, or uncertain. ECHO-PPI integrates weighted network topology, semantic protein profiles, and Gene Ontology evidence to identify evidence-potential nuclei, construct candidate modules, perform overlap-aware assignment, and export hierarchical confidence labels.

Evaluation on yeast protein-interaction data shows that ECHO-PPI preserves the behaviour of strong overlap-aware baselines while adding evidence-bundled auditability. Each protein-module assignment comes with topology, semantic, and Gene Ontology evidence scores plus a hierarchical confidence label, enabling curators to inspect, rank, and triage overlapping module predictions. Rather than claiming universal predictive superiority, ECHO-PPI makes overlapping protein-module predictions inspectable, confidence-aware, and reproducible for downstream biological interpretation.