Knowledge Graph-Driven AI in Biohealth: From Biomedical Discovery to Health Risk Prediction

Area of Science:

• Biohealth Informatics
• Artificial Intelligence
• Knowledge Representation

Background:

• Knowledge graphs (KGs) are pivotal for knowledge discovery in biohealth.
• Existing approaches often lack integration across molecular and population health levels.
• AI-driven methods are needed to unlock complex health insights.

Purpose of the Study:

• To present a framework for AI-driven discovery in biohealth using KGs.
• To demonstrate the framework's application in drug discovery and population health.
• To highlight the potential of open, interoperable knowledge networks.

Main Methods:

• KG construction, graph representation learning, and predictive modeling.
• Developed Protein Knowledge Network (ProKN) and KSMoFinder for kinase prediction.
• Built a Social Determinants of Health (SDoH) KG for suicide-risk prediction.

Main Results:

• KSMoFinder achieved state-of-the-art accuracy in predicting protein kinase associations.
• The SDoH KG model uncovered latent, multifactorial risk patterns for veteran suicide.
• Demonstrated KG-driven AI's ability to bridge molecular and population health studies.

Conclusions:

• KG-driven AI provides a reusable framework for accelerating biohealth discovery.
• This approach can address complex health challenges from molecular to population levels.
• Recommendations are provided for leveraging KGs in public health and translational research.