DNA-Diffusion is a Python library for generating synthetic regulatory DNA sequences using diffusion probabilistic models. It is designed for researchers and bioinformaticians working with genomics and synthetic biology, enabling the creation of cell-type-specific DNA elements for experimental validation or design.

How It Works

The project leverages diffusion probabilistic models, a class of generative models that learn to reverse a diffusion process (gradually adding noise) to generate new data. This approach allows for the generation of high-quality, realistic DNA sequences that capture the complex patterns found in regulatory elements. The model is trained on chromatin accessibility data to learn cell-type-specific sequence characteristics.

Quick Start & Requirements

• Install via uv sync after cloning the repository.
• Recommended: Linux with a recent GPU (e.g., A100).
• Compatible with CPU, but GPU is preferred for performance.
• Documentation: https://pinellolab.github.io/DNA-Diffusion

Highlighted Details

• Generates 200bp cell-type-specific synthetic regulatory elements.
• Provides scripts for both training and sequence generation.
• Supports debugging with a single-sequence training configuration.

Maintenance & Community

• Key contributors include Lucas Ferreira da Silva and Luca Pinello.
• Follows the all-contributors specification, welcoming contributions.

Licensing & Compatibility

• The repository does not explicitly state a license in the provided README.

Limitations & Caveats

• While compatible with CPU, performance is significantly better on a recent GPU.
• The README does not specify the exact license, which may impact commercial use or integration into closed-source projects.