OCKG: Ovarian Cancer Knowledge Graph

Explore an Ovarian Cancer/Immunology knowledge graph using natural language and GraphRAG patterns!

Comment

🎯 Inspiration

Immunotherapies offer hope for ovarian cancer, but the relationship is still under research. While working with researchers, I saw how knowledge graphs (KGs) could reveal new insights. Medical researchers, often not programmers, can benefit from natural language queries converted into Cypher queries.

🤖 What it does

OCKG provides a chatbot interface for querying ovarian cancer and immunology. Responses are informed by graph-based context, making them more interpretable and reliable.

🛠️ How we built it

  1. 📚 Semantic Medline → Neo4j: Importing medical literature relationships into a Neo4j graph.
  2. 🔗 Modus GraphQL API: Enabling seamless querying between the graph and frontend - the main function used was a prompt to query to execution function.
  3. 💻 Gradio Frontend: Simplifying natural language queries into actionable insights.

🚧 Challenges

  1. 📝 Text-to-Cypher: I used GPT-4o with context about the database schema to turn natural language queries into Cypher, but robustness needs testing. While Neo4j has a text2cypher API, I wanted to implement this functionality in Modus.
  2. 📈 Data Scale: Larger graphs were limited by Semantic Medline export constraints.

🏆 Accomplishments

  1. Integrated Semantic Medline into Neo4j for structured, queryable relationships.
  2. Built a GraphQL API with Modus for easy data access.
  3. Designed an intuitive Gradio frontend for researchers to explore the KG.

📚 What we learned

  1. Knowledge graphs are powerful for uncovering hidden relationships.
  2. Simplified interfaces are key for non-programmer researchers.
  3. Scaling large datasets requires efficient pipelines.
  4. AI-driven text-to-Cypher can streamline query workflows.

🚀 What's next

  1. 🔍 Bigger Graphs: Expand with full Semantic Medline data.
  2. ⚙️ Smarter Queries: Refine text-to-Cypher functionality for robustness.
  3. 📊 Predefined Templates: Create easy-to-use research queries.
  4. 🩺 Clinical Integration: Add real-world clinical data.
  5. 🌍 Open Source: Share OCKG with the research community for collaboration.

Built With

Share this project:

Updates