PharmaGraphX – Graph-Based Drug Discovery Platform

Revolutionizing drug discovery with AI. Our platform maps complex graph(drug-target) relationships, automates research with intelligent agents, and accelerates biomedical insights at scale

Comment 2

Inspiration

Drug discovery is slow, expensive, and inefficient. Major challenges include:

  • High Costs → Developing a single drug costs ~$1.1 billion on average.
  • Long Timelines → It takes 10–15 years to bring a drug from research to market.
  • High Failure Rates → Over 90% of drug candidates fail in clinical trials.
  • Data Complexity → Drug interactions, side effects, and biological pathways form a vast, interconnected system that is difficult to analyze using traditional methods.

What it does

PharmaGraph AI is an AI-powered drug discovery platform that accelerates research by leveraging:

  • Graph Representation → Captures relationships between drugs, proteins, diseases, and side effects for deeper insights.
  • Automated Querying → AI agents (AQL + cuGraph) simplify and accelerate data exploration.
  • GPU-Accelerated Analysis → Uses cuGraph for fast, large-scale computations.
  • Smart Data Filtering → Focuses on highly relevant connections, reducing noise and redundancy.
  • Scalable & Adaptive → Leverages ArangoDB, NetworkX, and cuGraph for efficient and scalable drug discovery.

How we built it

  • Implemented AI-powered graph-based analytics using ArangoDB, NetworkX, and cuGraph.
  • Developed three AI agents:
    • text_to_aql_to_text
    • text_to_cugraph_algorithm_to_text
    • text_to_table
  • Integrated natural language processing (NLP) with Langchain to interpret and execute hybrid queries.
  • Processed and optimized large datasets with over 24960 nodes and 395379 edges in ArangoDB.
  • Used publicly available drugs data from DrugCentral.
  • Ensured proper data preprocessing, traversal, and visualization using cuGraph.
  • Built an interactive UI for visualization and chatbot-based interactions.
  • Deployed the application on Heroku (Live Demo).

Challenges we ran into

  • Handling Large Datasets → Required efficient preprocessing and optimized database queries.
  • Graph Query Optimization → Ensured fast traversal and data retrieval using cuGraph and AQL.
  • Natural Query Execution → Developed AI agents to execute hybrid queries for graph traversal and data analysis.
  • Visualization Complexity → Built a user-friendly interface for intuitive exploration of drug relationships.

Accomplishments that we're proud of

  • Successfully built a scalable AI-powered drug discovery platform.
  • Integrated multiple AI agents for efficient graph query execution.
  • Processed and visualized massive pharmaceutical datasets.
  • Developed an interactive UI that enhances usability and accessibility.
  • Deployed a fully functional web application for real-world usage.

What we learned

  • Advanced graph-based drug discovery techniques using cuGraph and ArangoDB.
  • Optimized AI-driven query execution for large-scale datasets.
  • Enhanced natural language processing capabilities for structured query generation.
  • Improved scalability and deployment strategies for AI applications.

What's next for PharmaGraph AI

  • Expanding dataset integration to include real-time pharmaceutical research data.
  • Enhancing AI models for more accurate drug-target interaction predictions.
  • Developing advanced visualization tools for deeper data insights.
  • Optimizing GPU performance for faster query execution and analysis.
  • Exploring potential collaborations with pharmaceutical research institutions.

PharmaGraph AI transforms big pharmaceutical data into actionable insights, helping researchers discover new drugs faster and cheaper.

Built With

Share this project:

Updates