README.md

---
license: cc-by-nc-nd-4.0
tags:
- spatial-transcriptomics
- pathology
- histology
- deep-learning
- pytorch
---

# HistoPrism: Unlocking Functional Pathway Analysis from Pan-Cancer Histology via Gene Expression Prediction

<div align="center">

[![arXiv](https://img.shields.io/badge/arXiv-2601.21560-b31b1b.svg)](https://arxiv.org/abs/2601.21560)
[![Model Architecture](https://img.shields.io/badge/Model-Transformer-blue)]()
[![Dataset](https://img.shields.io/badge/Dataset-HEST-green)]()

</div>

**HistoPrism** is a deep learning model designed to bridge the gap between histology images (H&E) and spatial gene expression. 

Check out the details in the [github repo](https://github.com/susuhu/HistoPrism).

This repository contains the weights for the checkpoints in the paper trained on the HEST v1.1.0 dataset.

## 📄 Paper
**Title:** HistoPrism: Unlocking Functional Pathway Analysis from Pan-Cancer Histology via Gene Expression Prediction  
**Authors:** Hu, Susu and Zeng, Qinghe and Bhasker, Nithya and Kather, Jakob Nicholas and Speidel, Stefanie  
**Link:** [ICLR 2026 arXiv](https://arxiv.org/abs/2601.21560)


## 💻 Usage
To load this checkpoint, ensure you have the HistoPrism codebase or compatible model definition.

```python
from huggingface_hub import hf_hub_download
import torch

# Download the model checkpoint
checkpoint_path = hf_hub_download(repo_id="HuSusu/HistoPrism", filename="HistoPrism_split0.ckpt")

# Load weights (Pseudo-code: replace with your actual model class)
# model = HistoPrism(config=...)
# checkpoint = torch.load(path, map_location=map_location)
# model.load_state_dict(checkpoint["model_state"])