Bio-informatics AI agent for Drug discovery Research

Follow our wiki pages for more insight.

Overview

DEDA - think of ChatGPT for bioinformatics researchers.

It's designed as a lightweight, low-overhead tool that you can install in your laptop.

DEDA focuses on integrating workflows, helping researchers explore drug discovery, analyze proteins, and quickly test new ideas - all through simple commands in English.

For example, instead of writing code to pull protein data or run a molecule generator, you can ask:

Show known binding pockets for the SARS-CoV-2 virus.

The underlying Agentic AI does not do any guesswork. Instead it connects directly to real tools and data source, so answers are reliable and grounded.

The AI connects to trusted sources like UniProt, AlphaFold, OpenTargets and others. This grounds its answers in real biological and chemical data, cutting down on hallucinations (more details here). The result: bioinformatics workflows that are easier, more accurate, and repeatable—so researchers can focus on discovery instead of wrestling with tools.

Disclaimer: This tool helps portray and organize data, but it is not an inventor. Final insights and decisions must come from domain experts.

Collaboration Call: Contributions are needed to integrate advanced ML modeling components—such as Boltz-2—so the platform can generate and refine accurate protein data directly within our tool. For details, see our contribution guideline.

Quick Start

Installation

Create a virtual environment using python version 3.12 or later.

python3 -m venv venv

Then activate the environment:

source venv/bin/activate

Then install the required packages:

pip install -r requirements.txt

pip install -e .

Setting Up Environment Variables

Before running any component, configure the required environment variables. Copy the example environment file:

cp .env.example .env

Edit the .env file and add your OpenAI API key:

OPENAI_API_KEY=your_openai_api_key_here

---

Running the Application

This project provides multiple ways to interact with the bioinformatics AI agent:

1. Running MCP Server

The Model Control Protocol (MCP) server provides the core AI agent functionality that can be integrated with compatible clients like Claude Desktop.

Start the MCP server:

# Basic server
python -m drug_discovery_agent.interfaces.mcp.server

# With custom port
python -m drug_discovery_agent.interfaces.mcp.server --port 8081

The server will start on localhost:8080 by default and can be connected to by MCP-compatible clients.

Start the MCP Client:

python -m drug_discovery_agent.interfaces.mcp.client

2. Running Chat on CLI

For quick terminal-based interactions with the AI agent:

python -m drug_discovery_agent.chat
# or simply
chat

Debug and Verbose Mode

To see detailed tool selection and execution activity (useful for debugging or understanding how the AI agent works):

python -m drug_discovery_agent.chat --verbose
# or
python -m drug_discovery_agent.chat --debug

This will show:

• Tool selection decisions made by the LLM
• Tool execution steps and reasoning
• Input/output details for each tool call
• Agent's internal reasoning process

Try queries like: Show me details for UniProt ID P0DTC2, followed by What are the structural properties of this protein?

3. Running Chat on UI (Desktop App)

For a modern, user-friendly desktop experience, use the Electron-based chat interface:

📱 See the full Desktop App documentation →

Quick setup:

cd electron-app
npm install
npm run dev

The desktop app provides:

• Modern chat interface with message history
• Real-time AI responses with progress indicators
• Session management and conversation persistence
• Cross-platform support (Windows, macOS, Linux)
• Integrated backend server management

Creating Distribution Installers

To create installable packages for distribution:

📦 See detailed installer creation guide →

Quick installer build:

# Navigate to electron-app directory
cd electron-app

# Build Python backend first
npm run build:python

# Then build Electron installer
npm run build:mac    # or build:win, build:linux

Supports creating native installers for Windows (.exe), macOS (.dmg), and Linux (.AppImage) with automatic environment configuration and backend bundling.

4. Run the AI assistant using Docker

Pre-requisite Make sure your docker runs in rootless mode. If you can run

docker run hello-world

without sudo, you are good to go.

Change API Key Before Proceeding

Inside entrypoint.sh, replace the OPENAI_API_KEY, with your openAI api key.

ENV OPENAI_API_KEY=sk-proj-XXXX

Run the following command to build the docker image

docker build --no-cache -t mcp_app .

To test

docker run -it mcp_app

Enjoy chatting!