Hypothesizing mechanistic links between microbes and disease using knowledge graphs.

Area of Science:

• Biomedical informatics
• Microbiome research
• Systems biology

Background:

• The gut microbiome's link to disease is evident, but mechanisms remain unclear.
• Knowledge graphs offer a way to represent complex biomedical information.
• Understanding host-microbe interactions is crucial for disease etiology.

Approach:

• Constructed MGMLink, a knowledge graph linking microbes, genes, and metabolites.
• Employed semantically constrained shortest path search within the graph.
• Utilized a novel cosine similarity-based path prioritization for hypothesis generation.

Key Points:

• The knowledge graph approach successfully generates mechanistic hypotheses for host-microbe interactions.
• The method effectively explains observed relationships between gut microbes and disease phenotypes.
• Demonstrated applications in inflammatory bowel disease and Parkinson's disease.

Conclusions:

• Knowledge graphs provide a scalable and comprehensive method for generating mechanistic hypotheses in microbiome-disease research.
• This approach enhances the understanding of complex gut microbiome-disease relationships.
• Facilitates discovery of novel therapeutic targets and disease mechanisms.