README.md

metadata

pretty_name: STP-Bench
language:
  - en
license: cc-by-nc-sa-4.0
tags:
  - pathology
  - histopathology
  - spatial-transcriptomics
  - virtual-spatial-transcriptomics
  - spatial-gene-expression-prediction
  - whole-slide-image
  - computational-pathology
  - gene-expression
  - regression
  - benchmark
  - visium
  - xenium
size_categories:
  - 100K<n<1M

STP-Bench

STP-Bench is a comprehensive benchmark dataset for evaluating deep learning models on spatial transcriptomics prediction from histopathology whole-slide images (WSIs).

The dataset accompanies the paper introducing STP-Bench, a unified benchmark designed to standardize evaluation across multiple spatial transcriptomics technologies, tissue types, and experimental protocols.

Unlike previous studies that evaluate on a single cohort or technology, STP-Bench aggregates diverse publicly available spatial transcriptomics datasets into a unified benchmark, providing standardized data organization and evaluation protocols for robust and reproducible benchmarking.

---

📊 Dataset Statistics

STP-Bench consists of two benchmark settings:

• STP-Bench Internal: used for model development and benchmark evaluation.
• STP-Bench External: used to assess model generalization on independent cohorts.

The complete benchmark composition reported in the accompanying paper is summarized below.

Benchmark	Dataset	Source	Cancer Type	Platform	Slides	Patients	ST Spots
Internal	HEST-PRAD	HEST-1K	PRAD	Visium	23	23	62,710
	WUSTL-BRCA	WUSTL	IDC	Visium	47	21	157,138
	NCCHE-Xenium	NCCHE	LUAD	Xenium	16	16	34,519
	NCCHE-Visium	NCCHE	LUAD	Visium	43	43	96,954
	HEST-CCRCC	HEST-1K	RCC	Visium	24	24	74,220
	SMC-GBM	SNU	GBM	Visium	30	17	117,036
	WUSTL-PDAC	WUSTL	PAAD	Visium	19	14	66,447
External	🔒 MGB-PRAD	MGB	PRAD	Visium	8	8	31,438
	HEST-IDC	HEST-1K	IDC	Xenium	4	4	35,536
	Massey-TNBC	Massey	IDC	Visium	43	22	55,227
	HEST-LUNG	HEST-1K	LUAD	Xenium	2	2	5,206
	WUSTL-RCC	WUSTL	RCC	Visium	10	10	33,090
	HEST-GBM	HEST-1K	GBM	Visium	3	3	17,763
	HEST-PAAD	HEST-1K	PAAD	Xenium	3	3	7,571

Internal cohort: 202 slides, 158 patients, and 609,024 spatial transcriptomics spots.

External cohort: 73 slides, 52 patients, and 185,831 spatial transcriptomics spots.

🔒 MGB-PRAD is not included in this public Hugging Face release due to institutional data-sharing restrictions. The accompanying paper reports results on this in-house external cohort for additional generalization evaluation.

---

📁 Dataset Structure

STP-Bench/

├── metadata/
│   ├── dataset_metadata.json
│   ├── ...
│
├── wsis/
│   ├── slide_001.tif
│   ├── ...
│
├── patches/
│   ├── slide_001.h5
│   ├── ...
│
└── st/
    ├── slide_001.h5
    ├── ...

Directory	Description
metadata	Dataset metadata and slide annotations
wsis	Whole-slide H&E images
patches	Patch-level image features for each slide
st	Spatial transcriptomics measurements for each slide

---

💻 Supporting Code

Benchmark code is available at

GitHub

https://github.com/NEXGEM/STpredBench

The repository contains

• preprocessing pipeline
• data loading
• evaluation scripts
• baseline implementations
• benchmark splits
• training examples

---

⬇️ Loading Example

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="nexgem/STP-Bench",
    repo_type="dataset",
    local_dir="./STP-Bench"
)

📖 Citation

If you use STP-Bench, please cite

@article{stpbench2026,
  title={STP-BENCH: A Unified Systematic Benchmark for Virtual Spatial Transcriptomics from Histopathology Images},
  author={Chung, Youngmin and Ha, Ji Hun and Song, Andrew H. and Almagro-P{\'e}rez, Cristina and Seo, Chaeyoung and Suh, Won Jun and Beom, Jeong Won and Oh, Kyoung Bin and Ruppin, Eytan and Mahmood, Faisal and Lee, Joo Sang},
  journal={arXiv preprint arXiv:2609.05956},
  year={2026},
  doi={10.48550/arXiv.2609.05956}
}

---

📜 License

Please refer to the original licenses of each constituent dataset.

The benchmark organization and preprocessing pipeline are released under the CC BY 4.0 License unless otherwise specified.

---

📬 Contact

Youngmin Chung

• Sungkyunkwan University, Suwon, Republic of Korea
• ymblue@skku.edu

Ji Hun Ha

• Sungkyunkwan University, Suwon, Republic of Korea
• gkwlgns323@skku.edu