Analysis of Pollutant Exposure and Toxicological Risk in Aquatic Species Using Knowledge Graphs with Louvain Community Detection

Abstract

Heavy metal pollution poses significant risks to biodiversity and food safety. We developed a novel framework integrating Knowledge Graph (KG) technology with the Louvain community detection algorithm to identify multi-metal contamination patterns and assess cumulative toxicological risks in aquatic species. We constructed a comprehensive KG from 310 scientific papers containing 55,297 contamination records across 778 species and 559 pollutants in China. The KG models entities including heavy metals, aquatic species, and toxic effects, with relationships such as 'bioaccumulates_in' and 'poses_health_risk'. Risk quantification aggregates evidence paths within the KG, weighted by toxicity metrics including Target Hazard Quotient (THQ) and Incremental Lifetime Cancer Risk (ILCR). The Louvain algorithm identified 23 densely connected communities with exceptionally high modularity (Q = 0.847), indicating strong community structure where clusters represent distinct multi-metal contamination syndromes. Community 1, the largest cluster (187 nodes), revealed mercury, lead, cadmium, and chromium co-contamination affecting 45 species with mean Total Target Hazard Quotient (TTHQ) = 6.34, predominantly impacting carnivorous species in industrial coastal regions. The identified high-risk clusters aligned with known pollution hotspots in independent field studies. This methodology transforms traditional single-contaminant assessment into systemic, multi-pollutant risk evaluation, providing actionable intelligence for targeted environmental monitoring, evidence-based consumption guidelines, and policy intervention strategies.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint