Context-Aware Multi-Agent AI System to Explore the Oxidative Stress in Cardiovascular Medicine

Overview

This repository contains the implementation of a context-aware multi-agent AI system designed to explore the complex interplay between oxidative stress (OS) and cardiovascular diseases (CVDs). The system leverages advanced AI methodologies, including knowledge graph (KG) construction, graph neural networks (GNNs) for link prediction, and a modular multi-agent framework to dynamically validate and refine insights. The objective is to bridge the gap between fragmented biomedical data and actionable discoveries, accelerating research in cardiovascular medicine.

---

Features

• Comprehensive Knowledge Graph (KG):

  • Integrates biomedical data from PubMed, UniProt, DrugBank, and Reactome.
  • Models nodes (proteins, pathways, drugs, diseases) and edges (relationships) with high fidelity.

• Graph Neural Network (GNN):

  • Implements state-of-the-art GNN models for predicting novel OS-CVD relationships.
  • Identifies high-confidence links between biomarkers, pathways, and drug targets.

• Multi-Agent AI Framework:

  • Modular architecture with specialized agents:
    • UniProt Agent: Protein data and functional annotations.
    • CVD Agent: Pathways and mechanisms underlying cardiovascular diseases.
    • OS Agent: Analysis of oxidative stress biomarkers and mechanisms.
    • Drug Agent: Drug interactions and therapeutic implications.
    • Reactome Agent: Systems-level analysis of metabolic and signaling pathways.
  • Central orchestrator for task management and inter-agent communication.

• Dynamic Analysis and Refinement:

  • Agents leverage contextual understanding to refine predictions.
  • Feedback loops ensure iterative improvement of outputs.

• Interactive Insights Visualization:

  • Visualizes KGs and predicted relationships via interactive dashboards.

---

Installation

Prerequisites

• Python 3.8 or later
• Recommended: Google Cloud Platform (GCP) account for deploying scalable workflows.
• Required Python Libraries:
  • tensorflow, torch
  • neo4j, networkx
  • flask, fastapi
  • numpy, pandas
  • matplotlib, seaborn, plotly
  • google-cloud-* for GCP integrations

Setup

1. Clone this repository:

   git clone https://github.com/your-username/context-aware-ai-system.git
   cd context-aware-ai-system

2. Set up a Python virtual environment:

   python -m venv venv
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Google Cloud services:

   • Set up a Google Cloud project and enable APIs for Vertex AI, BigQuery, Cloud Storage, and Compute Engine.
   • Authenticate using the gcloud CLI:

     gcloud auth login
     gcloud config set project [PROJECT_ID]

5. Start the application locally:

   python app.py

---

Usage

Knowledge Graph Construction

• Run scripts/build_knowledge_graph.py to construct the KG from biomedical datasets:

  python scripts/build_knowledge_graph.py --input data/ --output kgraph/

GNN Model Training

• Use scripts/train_gnn.py to train the link prediction model:

  python scripts/train_gnn.py --graph kgraph/ --output models/

Multi-Agent System

• Launch the multi-agent framework with:

  python multi_agent_system.py

Visualization

• Visualize insights via interactive dashboards:

  python visualization.py

---

Project Structure

context-aware-ai-system/
├── data/                   # Raw and preprocessed datasets
├── kgraph/                 # Knowledge graph files
├── models/                 # Trained GNN models
├── scripts/                # Scripts for data processing, KG construction, and GNN training
├── agents/                 # Implementation of specialized agents
├── visualization/          # Scripts for visualizing insights
├── app.py                  # Main entry point for running the system
├── requirements.txt        # Python dependencies
└── README.md               # Project documentation

---

Contributing

We welcome contributions to this project! To contribute:

1. Fork this repository.
2. Create a new branch for your feature or bug fix:

   git checkout -b feature-name

3. Commit your changes:

   git commit -m "Add a meaningful commit message"

4. Push to your fork and submit a pull request.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Acknowledgments

This work is inspired by:

• Panday et al., "Context-aware Multi-agent AI System to Explore the Oxidative Stress in Cardiovascular Medicine" (2024).
• Panday et al., "Data-Driven Insights into the Association Between Oxidative Stress and Calcium-Regulating Proteins in Cardiovascular Disease" (2024).
• Google Cloud's Vertex AI and Neo4j for advanced data integration and processing.

---

Contact

For questions or collaboration opportunities, please reach out to:

• Namuna Panday | Department of Physiology, UCLA | namuna@ucla.edu
• Dibakar Sigdel | Department of Physiology, UCLA | dibakarsigdel@ucla.com