Knowledge Graph Extraction from Biomedical Literature for Alkaptonuria Rare Disease

A research consortium of 14 scientists has published a novel AI methodology for constructing biomedical knowledge graphs (KGs) specifically for Alkaptonuria (AKU), an ultra-rare autosomal recessive metabolic disorder. The disease, caused by mutations in the HGD gene, leads to a pathological accumulation of homogentisic acid and results in severe systemic complications like premature spondyloarthropathy and cardiovascular issues. Due to its rarity, AKU is critically underrepresented in existing biomedical knowledge bases, creating a significant gap in research connectivity.

The team's approach centers on a large-scale text-mining pipeline built on PubTator3, a powerful biomedical concept recognition tool. This AI framework extracts structured relationships from scientific literature to build two distinct knowledge graphs of varying complexity. After construction, the researchers validated the graphs against established biochemical knowledge, ensuring scientific accuracy. The final KGs successfully mapped genes, diseases, and therapies potentially related to AKU, uncovering systemic disease interactions and highlighting novel therapeutic targets that were previously obscured by data scarcity.

This work demonstrates a scalable computational framework that can be adapted to other rare diseases. By transforming fragmented literature into interconnected knowledge graphs, the methodology helps researchers visualize complex disease mechanisms and identify hidden relationships. The success with Alkaptonuria suggests that similar AI-driven approaches could accelerate discovery for hundreds of other rare conditions that suffer from the 'data desert' problem, where limited patient information hinders traditional research pathways.