Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching

This is our PyTorch implementation for the paper:

Tinglin Huang, Tianyu Liu, Mehrtash Babadi, Wengong Jin, and Rex Ying (2025). Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching. Paper in arxiv.

We recently extended STFlow and released STPath, a generative pretrained model capable of directly predicting the expression levels of 38,984 genes from histology images without further fine-tuning. Feel free to check out the paper and code, which includes an easy-to-use API.

Usage

We are currently cleaning and organizing the codebase and will release it soon.

Reference

If you find our work useful in your research, please consider citing our paper:

@inproceedings{huang2025stflow,
  title={Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching},
  author={Huang, Tinglin and Liu, Tianyu and Babadi, Mehrtash and Jin, Wengong and Ying, Rex},
  booktitle={International Conference on Machine Learning},
  year={2025}
}

@article{huang2025stpath,
  title={STPath: A Generative Foundation Model for Integrating Spatial Transcriptomics and Whole Slide Images},
  author={Huang, Tinglin and Liu, Tianyu and Babadi, Mehrtash and Ying, Rex and Jin, Wengong},
  journal={bioRxiv},
  pages={2025--04},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}