Repuragent - An AI Scientist for Drug Repurposing

Overview

Drug repurposing offers an efficient strategy to accelerate therapeutic discovery by identifying new indications for existing drugs. However, the process remains hindered by the heterogeneity of biological and chemical data and the difficulty of forming early, evidence-based hypotheses about candidate drugs, targets, and clinical endpoints. We introduce Repuragent (Drug Repurposing Agentic System), a proof-of-concept multi-agent framework designed to autonomously plan, execute, and refine data-driven repurposing workflows under human-in-the-loop supervision. The system integrates autonomous research, data extraction, knowledge graph (KG) construction, and analytical reasoning with an adaptive long-term memory mechanism that improves the system over time.

Version Announcement:

This repository contains a local application that you can clone and run on your own machine. As a result, all data files, database and chat memory remain on your device. Additionally, we offer several alternative versions:

• Repuragent Web: This is the web version of Repuragent, which can be used without installation. Simply register for an account to access the app.
• Repuragent Demo Page: This is a static demo of Repuragent designed for viewing purposes only. You cannot interact with the agent in this mode.
• Documentation: User guides and technical details.

Core Agent Architecture

• Planning Agent: Decomposes complex tasks using episodic memory, Standard Operating Procedures (SOPs), and academic publications.
• Supervisor Agent: Delegate tasks to specialized agents and track the completion status.
• Research Agent: Performs literature mining, accesses knowledge graphs, and integrates biomedical databases.
• Prediction Agent: Executes molecular property predictions using pre-trained ML models.
• Data Agent: Manages multi-format data processing, SMILES standardization, and visualization
• Report Agent: Generates comprehensive reports.

Advanced Memory Systems

• Episodic Memory: Pattern extraction from successful executions to improve future planning
• Short-term Memory: SQLite-based conversation persistence with thread management
• SOP RAG System: Retrieval-augmented generation using professional Standard Operating Procedures

Quick Start

Prerequisites

• Docker Desktop
• OpenAI API key from platform.openai.com
• (Optional) LangSmith account for tracing from smith.langchain.com

Initial Setup

# 1. Clone repository
git clone https://github.com/your-username/repuragent.git
cd repuragent

# 2. Create .env file with required API keys
## Mandatory API key
echo "OPENAI_API_KEY=your-openai-api-key-here" > .env   

## Optional set-up
echo "LANGCHAIN_TRACING_V2=true" >> .env
echo "LANGCHAIN_ENDPOINT=https://api.smith.langchain.com" >> .env
echo "LANGCHAIN_API_KEY=your-langsmith-api-key-here" >> .env
echo "LANGCHAIN_PROJECT=repuragent" >> .env

# 3. Build and run Docker containers
docker-compose up --build

Open http://localhost:7860 to access the Gradio application.

Daily Usage

# Start the application (after initial setup)
docker-compose up

# Stop the application
docker-compose down

# View logs
docker-compose logs -f

LangSmith Setup (Optional)

1. Create account at smith.langchain.com
2. Get your API key from the settings page
3. Add the LangSmith variables to your .env file as shown above
4. Restart Docker containers to apply changes

Project Structure

repuragent/
├── app/           # Gradio UI interface
├── core/          # AI agents and logic
├── backend/       # Memory, tools, and RAG systems
├── models/        # Pretrained ML models
├── data/          # Input data will be stored here
├── results/       # Output files will be stored here
└── main.py        # Entry point