Skip to content

Getting Started with DNALLM

Welcome to DNALLM! This guide will walk you through the initial setup and first steps with this powerful toolkit for DNA language models.

1. Project Overview

DNALLM is an open-source toolkit designed for large language model (LLM) applications in DNA sequence analysis and bioinformatics. It provides a comprehensive suite for:

  • Model Training & Fine-tuning: Supports a variety of DNA-related tasks, including classification, regression, and named entity recognition (NER).
  • Inference & Benchmarking: Enables efficient model inference, mutagenesis analysis, and multi-model benchmarking.
  • Data Processing: Includes tools for dataset generation, cleaning, formatting, and augmentation.
  • Model Management: Offers flexible loading of different DNA language models.
  • Extensibility: Features a modular design for easy integration and secondary development.

2. Quick Start: Installation

Getting DNALLM installed is the first step. We recommend using uv, a fast Python package manager, within a virtual environment.

Prerequisites

  • Python 3.10 or higher
  • Git
  • A virtual environment manager like venv (built-in) or conda.

Installation Steps

  1. Clone the Repository 

    git clone https://github.com/zhangtaolab/DNALLM.git
cd DNALLM

  2. Create and Activate a Virtual Environment We'll use Python's built-in venv. 

    # Create the environment
python -m venv .venv

# Activate it
source .venv/bin/activate  # On Linux/macOS
# .venv\Scripts\activate  # On Windows

  3. Install uv and DNALLM 

    # Install uv, the fast package manager
pip install uv

# Install DNALLM and its core dependencies
uv pip install -e '.[base]'

  4. Verify the Installation 

    python -c "import dnallm; print('DNALLM installed successfully!')"

GPU and Mamba Support (Optional)

For accelerated performance, you can install support for GPU and specialized model architectures like Mamba.

# For GPU support with CUDA 12.4
uv pip install -e '.[cuda124]'

# For native Mamba architecture support
uv pip install -e '.[mamba]' --no-cache-dir --no-build-isolation

For more detailed instructions, please see the full Installation Guide.

3. Basic Usage: First Inference

Let's run a simple inference to see the toolkit in action. This example loads a pre-trained model and uses it to make a prediction on a DNA sequence.

from dnallm import load_config, load_model_and_tokenizer
from dnallm.inference import DNAInference

# 1. Load a pre-defined configuration
configs = load_config("./example/notebooks/inference/inference_config.yaml")

# 2. Load a pre-trained model and its tokenizer from Hugging Face
model_name = "zhangtaolab/plant-dnagpt-BPE-promoter"
model, tokenizer = load_model_and_tokenizer(
    model_name, 
    task_config=configs["task"], 
    source="huggingface"
)

# 3. Initialize the inference engine
inference_engine = DNAInference(config=configs, model=model, tokenizer=tokenizer)

# 4. Make a prediction on a sample sequence
sequence = "TCACATCCGGGTGAAACCTCGAGTTCCTATAACCTGCCGACAGGTGGCGGGTCTTATAAAACTGATCACTACAATTCCCAATGGAAAAA"
inference_result = inference_engine.infer(sequence)

print(f"Inference result: {inference_result}")

You've just completed your first task with DNALLM! Now you're ready to explore more complex workflows.