---
pretty_name: "Latent DNA Diffusion Sequences (hg19)"
tags:
- genomics
- biology
- generative-ai
- diffusion-models
- dna
- huggingscience
- science
license: "other" 
task_categories:
- text-generation # Generating DNA sequences is analogous to text generation
size_categories:
- 10M<n<100M
---

# Dataset Card for Latent DNA Diffusion

## Dataset Description

This dataset contains a collection of human DNA sequences, processed for the purpose of training generative models like the one described in the **Latent DNA Diffusion** project.  

- **Source**: Human reference genome assembly **hg19 (GRCh37)**  
- **Sequence length**: 256 base pairs  
- **Format**: HDF5 file containing uniformly processed sequences  

The primary purpose of this dataset is to serve as a training corpus for models that can learn the underlying patterns of the human genome and generate novel, realistic DNA sequences.  

### Applications
- Data augmentation for genomics  
- Studying gene regulation  
- Generating synthetic genomic data to preserve patient privacy  

---

## How to Use

The data is stored in a single **HDF5** file.  

Example usage in Python:

```python
import h5py
from huggingface_hub import hf_hub_download

# Download the HDF5 file from Hugging Face Hub
file_path = hf_hub_download(
    repo_id="Zehui127127/latent-dna-diffusion",
    filename="human_hg19_256.hdf5",
    repo_type="dataset"
)

# Open and explore the file
with h5py.File(file_path, 'r') as f:
    print("Available keys:", list(f.keys()))
    sequences = f['sequences'][:]  # Example key
    print("Shape of dataset:", sequences.shape)