Bridging AI and biology: Inside the AWS and NVIDIA Open Data knowledge graph hackathon

This article describes an AWS and NVIDIA hackathon where 50 researchers developed AI solutions combining knowledge graphs with graph-based retrieval-augmented generation (GraphRAG) for biomedical research.

• Seven teams created prototypes integrating genomics, proteomics, and literature data across siloed databases
• GeNETwork: Multi-scale cancer genomics and pharmacology knowledge graph for precision oncology
• ECoGraph: Integrated genomic and proteomic data identifying colorectal cancer biomarkers
• ClassiGraph: Multi-omics knowledge graph with graph neural networks for cancer classification
• EasyGiraffe: Simulator-based validation framework for polygenic variant extraction
• MIDAS: Modular knowledge graph system with natural language LLM interface for discovery
• KG-LLM Garbage Collection: AI-assisted tool identifying and pruning erroneous graph edges
• BioGraphRAG: Bridges biomedical data with citation-supported question answering
• All projects open-source with transparent documentation and reproducible methodologies

The hackathon demonstrated how knowledge graphs and GraphRAG create trustworthy, evidence-grounded AI systems for biomedical research with scalable cloud-native architecture.