SpaTranslator

• SpaTranslator is a graph-based generative model for cross-modality prediction in spatial omics data. It enables the in silico generation of missing omics modalities—such as spatial ATAC-seq, RNA-seq, and histone modification profiles—across paired or unpaired tissue sections, by leveraging spatial neighborhood information and adversarial learning.

🧠 Highlights

• 🔄 Cross-modality generation between spatial transcriptomics and epigenomics/proteomics
• 📈 Superior performance on 12 benchmark datasets over state-of-the-art methods (multiDGD, JAMIE, scButterfly, scPair)
• 🧬 Supports multiple data types: RNA, ATAC, histone marks (H3K27me3, H3K4me3), protein
• 📍 Graph Neural Network integration for spatial structure modeling
• 🧪 Flexible in intra- and inter-dataset prediction tasks

📂 Datasets

SpaTranslator is benchmarked on:

• Spatial-ATAC-RNA-seq (mouse brain)
• MISAR-seq (mouse brain at stages E11, E13, E15, E18)
• 10x Visium RNA-Protein (human tonsil&lymph node)

Detailed documentation is written in the jupyter notebook

🚀 Installation Guide

Install with pip

conda create -n SpaTranslator python==3.9
pip install spatranslator

Install PyTorch Geometric and Related Dependencies It's crucial to install versions of these packages that match your PyTorch and CUDA versions: https://pytorch-geometric.readthedocs.io/en/latest/install/installation.html#installation-from-source

The following command installs pyg_lib, torch_scatter, torch_sparse, torch_cluster, and torch_spline_conv. Note: The URL https://data.pyg.org/whl/torch-2.6.0+cu124.html specifies wheels for PyTorch 2.6.0 and CUDA 12.4. Adjust ${TORCH} and ${CUDA} in the URL below to match your installed PyTorch version and CUDA version. You can find the correct URL on the PyTorch Geometric installation page.

# Example for PyTorch 2.6.0 and CUDA 12.4
# YOU MUST ADJUST THE URL BASED ON YOUR PYTORCH AND CUDA VERSION
pip install torch_geometric
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.6.0+cu124.html

📄 License

SpaTranslator is released under the MIT License.
You are free to use, modify, and distribute this software for academic and non-commercial purposes with proper attribution.

📬 Contact

If you have questions, feature requests, or would like to contribute:

• 💡 Issues: Please submit via the GitHub Issues Page
• ✉️ Email: maosheng@westlake.edu.cn/donghongyu@westlake.edu.cn