🧬 LLM4EVO2 Intelligent Sequence Design Platform

🚀 Cell-free System Efficient Expression Regulatory Sequence Design Tool Based on NVIDIA EVO2-40B and Zhipu GLM-4.5

Integrates three-stage design workflow, LLM Agent intelligent optimization system, and natural language interaction interface to provide bioengineers with powerful DNA sequence design capabilities

📋 Overview

This project implements a Python program based on the interaction between deep learning model EVO2 and large language model GLM-4.5-x, specifically designed for efficient expression regulatory sequence design in cell-free systems. Through a three-stage iterative optimization process, it achieves intelligent generation, validation, and optimization of DNA sequences, ultimately producing functional genetic regulatory elements.

✨ Features

• Three-stage Iterative Design Process: Unconstrained Exploration → Constrained Generation → Modular Validation
• Dual Model Collaboration: NVIDIA EVO2-40B + Zhipu GLM-4.5-x
• LLM Agent Automatic Optimization System: Intelligent Agent collaborative sequence optimization system
• Professional Sequence Analysis: GC content, functional element identification, secondary structure prediction
• Cell-free System Optimization: Specialized sequence optimization for cell-free expression systems
• GFP Expression Experiment Demo: Complete GFP expression regulatory sequence design workflow
• Interactive User Interface: Beautiful command-line interface and menu system based on Rich
• FASTA Sequence Export: Support for standard format sequence file export
• Natural Language Interaction: Multi-turn dialogue for sequence design requirements
• Agent Intelligent Optimization: Quality-driven automatic iteration and parameter adjustment

📦 Prerequisites

Install uv (Python Package Manager)

macOS:

# Using Homebrew (recommended)
brew install uv

# Or using curl
curl -LsSf https://astral.sh/uv/install.sh | sh

Linux:

# Using curl
curl -LsSf https://astral.sh/uv/install.sh | sh

# Or using pip
pip install uv

Windows:

# Using PowerShell
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

# Or using pip
pip install uv

🚀 Installation

# Clone the project
git clone https://github.com/SmartisanNaive/LLM_for_EVO2.git
cd LLM_for_EVO2

# Create virtual environment and install dependencies using uv
uv sync

# Install project to virtual environment
uv pip install -e .

⚙️ Configuration

API Setup

# Configure API keys
uv run evo2-designer setup

# Manage API configuration
uv run evo2-designer api-config

# Test API connection
uv run evo2-designer test

Configuration Files

• API Configuration: api_config.json (project root)

Example configuration:

{
  "nvidia_api_key": "your_nvidia_api_key_here",
  "glm_api_key": "your_glm_api_key_here",
  "evo2_base_url": "https://health.api.nvidia.com/v1/biology/nvidia/evo",
  "glm_base_url": "https://open.bigmodel.cn/api/paas/v4/"
}

🧬 Usage

Interactive Interface (Recommended)

# Launch interactive menu
uv run python main.py
# or
uv run evo2-designer interactive

Command Line Interface

# Basic commands
uv run evo2-designer setup                    # Initialize configuration
uv run evo2-designer test                     # Test API connection
uv run evo2-designer design --prompt "TAATACGACTCACTATAGGG" --length 99
uv run evo2-designer analyze --sequence "ATCGATCGATCG"
uv run evo2-designer list-projects            # View project list
uv run evo2-designer agent-config             # Configure Agent parameters

GFP Expression Experiment

# Run GFP experiment demo
uv run evo2-designer gfp-demo --max-length 140 --target-length 120 --export

Designs DNA regulatory sequences for GFP protein expression in cell-free systems, including T7 promoter, 5'UTR with RBS, and start codon ATG.

Programming Interface

from evo2_sequence_designer import (
    Evo2Client, GLMClient, ThreeStageDesigner,
    DesignParameters
)

# Initialize clients
evo2_client = Evo2Client(Evo2Config(api_key="your_nvidia_api_key"))
glm_client = GLMClient(GLMConfig(api_key="your_glm_api_key"))

# Create designer
designer = ThreeStageDesigner(evo2_client, glm_client)

# Run design
parameters = DesignParameters(
    initial_prompt="TAATACGACTCACTATAGGG",
    target_length=99
)
project = designer.run_complete_design(parameters)
print(f"Final sequence: {project.final_sequence}")

📚 API Reference

EVO2Client

from evo2_sequence_designer.models.evo2_client import Evo2Client, Evo2Config

config = Evo2Config(api_key="your_nvidia_api_key")
client = Evo2Client(config)
response = client.generate_sequence(
    prompt="TAATACGACTCACTATAGGG",
    max_tokens=100,
    temperature=0.7
)

GLMClient

from evo2_sequence_designer.models.glm_client import GLMClient, GLMConfig

config = GLMConfig(api_key="your_glm_api_key")
client = GLMClient(config)
analysis = client.analyze_sequence(
    sequence="ATCGATCGATCG",
    analysis_type="optimization"
)

SequenceAnalyzer

from evo2_sequence_designer.analysis import SequenceAnalyzer

analyzer = SequenceAnalyzer()
analysis = analyzer.analyze_sequence("ATCGATCGATCG")
print(f"Quality score: {analysis.quality_score}")

🛠️ API Services

• NVIDIA EVO2: https://build.nvidia.com/arc/evo2-40b (DNA sequence generation)
• Zhipu GLM: https://open.bigmodel.cn/ (Biological analysis)

📁 Project Structure

evo2-sequence-designer/
├── main.py                     # Program entry
├── src/evo2_sequence_designer/
│   ├── main.py                 # Main module
│   ├── models/                 # Model interfaces
│   │   ├── evo2_client.py     # EVO2 client
│   │   └── glm_client.py      # GLM client
│   ├── analysis/              # Sequence analysis
│   ├── design/                # Design workflow
│   ├── agents/                # Agent system
│   └── demos/                 # Demo modules
├── api_config.json            # API configuration
└── pyproject.toml

🔧 Development

Requirements

• Python 3.11+
• uv package manager
• NVIDIA API access
• Zhipu AI API access

Dependencies

• biopython: Bioinformatics analysis
• requests: HTTP requests
• zhipuai: Zhipu AI SDK
• rich: CLI interface
• typer: Command line framework

Extension Development

Adding New Agent

Extend BaseAgent class in agents/ directory to implement custom optimization logic.

Custom Analysis

Extend SequenceAnalyzer in analysis/ directory for new sequence analysis features.

New Models

Add new model clients in models/ directory following the existing interface pattern.

Command Line Extensions

Add new commands using the typer framework in the main application.

📈 Performance

• API call caching
• Batch processing
• Intelligent retry mechanism
• Asynchronous requests

🤝 Contributing

1. Fork the project
2. Create a feature branch
3. Commit changes
4. Submit a Pull Request

📄 License

MIT License. See the LICENSE file for details.

🙏 Acknowledgments

• NVIDIA: EVO2-40B model API
• Zhipu AI: GLM-4.5-x model support
• BioPython: Bioinformatics tools

---

Note: Valid API keys required. Please comply with service terms.