Skip to content

repolevedmaster/QuantumDrug-Agent

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 

Repository files navigation

🧬 Quantum-Agent Drug Discovery Platform

Next-generation AI-powered drug discovery platform based on Multi-Agent systems

LLM + Tensor Networks (MPS) + Molecular Generation + Docking + ADMET + Clinical Ranking


Overview

The Quantum-Agent Drug Discovery Platform is an automated AI system that takes a disease as input and performs:

  • Scientific literature retrieval
  • Target protein inference
  • Therapeutic hypothesis generation
  • Molecular candidate generation
  • Tensor Network (MPS)-based search space compression
  • Molecular docking simulation
  • ADMET evaluation
  • Clinical success prediction
  • Final candidate ranking and reasoning generation

All processes are connected in a unified pipeline and executed through a Streamlit-based GUI.


Features

Research Agent

  • PubMed literature search
  • Extraction of recent research papers
  • Automatic target protein inference
  • RCSB PDB structure retrieval
  • Automatic receptor acquisition

Hypothesis Agent

Based on disease context and literature:

  • Therapeutic strategy generation
  • Mechanism of action inference
  • LLM-based hypothesis generation

Generator Agent

Generates molecular candidates using:

  • Seed molecules
  • Mutation operations
  • Crossover operations
  • RDKit-based molecular generation

Quantum Agent (Tensor Network)

This is the core innovation of the system.

Molecules are transformed into:

Fingerprint → Tensor → Matrix Product State (MPS)

This representation is used to compute:

  • Bond entanglement
  • Entropy

Using this, the system reduces a large molecular search space by selecting only the most informative candidates.


Docking Agent

Supports two modes:

AutoDock Vina

Performs real molecular docking simulation

Heuristic Docking

Fallback mode used when Vina/OpenBabel is unavailable:

  • RDKit descriptor-based scoring

ADMET Agent

Two evaluation modes:

ChemBERTa

Uses a pretrained molecular language model

RDKit Descriptor-Based Evaluation

  • Lipinski Rule of Five
  • QED (drug-likeness score)
  • Physicochemical descriptors

Clinical Agent

Ranks final candidates using:

  • Binding score
  • ADMET score
  • Drug-likeness
  • Clinical success prediction score

Reasoning Agent

Uses an LLM to generate scientific explanations for:

  • Why the selected molecule is optimal
  • How it aligns with the therapeutic hypothesis
  • Mechanistic justification

Architecture

Disease
   │
   ▼
Research Agent
   │
   ▼
Hypothesis Agent
   │
   ▼
Generator Agent
   │
   ▼
Quantum MPS Agent
   │
   ▼
Docking Agent
   │
   ▼
ADMET Agent
   │
   ▼
Clinical Agent
   │
   ▼
Reasoning Agent
   │
   ▼
Markdown Report

Technology Stack

AI

  • Ollama
  • Llama 3.1
  • LangChain

Molecular AI

  • RDKit
  • ChemBERTa

Quantum-Inspired Computing

  • Quimb
  • Matrix Product State (MPS)
  • Tensor Networks

Drug Discovery Tools

  • AutoDock Vina
  • OpenBabel
  • PubMed API
  • RCSB PDB

Visualization

  • Streamlit
  • Plotly

Installation

git clone https://github.com/repolevedmaster/QuantumDrug-Agent

cd Quantum-Agent-DrugDiscovery

Install dependencies:

pip install -r requirements.txt

Download Ollama model:

ollama pull llama3.1

Run the application:

streamlit run app.py

Example Pipeline

Disease

↓

Research

↓

Target Protein

↓

Hypothesis

↓

Candidate Molecules

↓

Tensor Network Compression

↓

Docking

↓

ADMET

↓

Clinical Ranking

↓

Top Drug Candidate

↓

Scientific Reasoning

↓

Markdown Report

Screenshots

To be added later

  • Dashboard
  • MPS Visualization
  • Candidate Ranking
  • Report View

Future Work

  • Graph Neural Networks for molecule generation
  • Diffusion models for molecular design
  • Reinforcement learning optimization
  • AlphaFold structure integration
  • Multi-target drug discovery
  • Protein language model integration
  • Automated binding site detection
  • GPU-accelerated tensor networks

Project Structure

.
├── app.py
├── requirements.txt
├── README.md
└── report.md

License

MIT License


Disclaimer

This project is developed for research and educational purposes only.

The generated compounds are not actual drugs and must not be used for clinical or medical purposes.


Author

Quantum-Agent Drug Discovery Platform

Developed with:

  • Python
  • Streamlit
  • LangChain
  • RDKit
  • Quimb
  • ChemBERTa
  • AutoDock Vina
  • Ollama

About

LLM과 Tensor Network(MPS)를 활용한 멀티 에이전트 기반 신약 후보 탐색 플랫폼

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages