Biased Random Walk on a Multiscale Interactome Prioritizes Candidate Herbs and Active Constituents for Psoriatic Arthritis

Abstract

Psoriatic arthritis (PsA) is a chronic immune-mediated arthropathy characterized by synovitis, enthesitis, and coupled bone erosion and pathological new-bone formation. Although biologics have improved outcomes, incomplete or lost response and safety constraints motivate the search for complementary multi-target candidates. Here, herb–compound records from OASIS; compound–target interactions from DrugBank, TTD, and STITCH; and curated PsA disease genes from DisGeNET were integrated into a multiscale interactome, and a biased random walk with restart was applied to prioritize herbs and constituents. Herbs were ranked by diffusion-profile similarity to PsA together with disease–target overlap. The prioritized set comprised three literature-supported comparator herbs and six candidate herbs without prior PsA evidence. Enrichment of the shared targets against PsA-relevant KEGG pathways returned IL-17 signaling, TNF signaling, Th17-cell differentiation, and osteoclast differentiation, while integrated network analyses—rather than pathway enrichment alone—identified TRAF3IP2 and RUNX2, together with BMP4, as recurrent nodes linking inflammation to bone remodeling. Compound-resolved subnetworks nominated constituent-level mechanisms, and site-resolved docking identified α-asarone as the most ligand-efficient molecule within the tested candidate panel and, provisionally, iNOS as the target with the most consistently favorable ligand efficiency in that panel. These convergent computational findings are hypothesis-generating and define candidates for experimental evaluation.