miniMTI-CRC Example Data

Example single-cell imaging data for testing miniMTI, a molecularly anchored virtual staining framework for multiplex tissue imaging panel reduction.

Paper: bioRxiv 2026.01.21.700911
Code: GitHub
Model: changlab/miniMTI-CRC

Dataset Description

10,000 single-cell image patches randomly sampled (seed=42) from CRC-Orion sample CRC04 (colorectal cancer tissue WSI, RareCyte Orion platform).

File

• example_CRC04_10k.h5 — HDF5 file (~178 MB)

HDF5 Structure

Channel Ordering (20 raw channels)

Channels 0–16 are immunofluorescence markers. Channels 17–19 are co-registered H&E RGB. The miniMTI model treats each IF channel as a separate marker and the three H&E channels as a single marker (18 markers total).

Usage

from huggingface_hub import hf_hub_download

# Download example data
path = hf_hub_download(
    repo_id="changlab/miniMTI-CRC-example",
    filename="example_CRC04_10k.h5",
    repo_type="dataset",
)

# Run inference with miniMTI
python scripts/inference_example.py \
    --val-file $path \
    --input-channels 17,6,11,13

Citation

@article{sims2026minimti,
  title={miniMTI: minimal multiplex tissue imaging enhances biomarker expression prediction from histology},
  author={Sims, Z. and Govindarajan, S. and Ait-Ahmad, K. and Ak, C. and Kuykendall, M. and Mills, G. B. and Eksi, E. and Chang, Y. H.},
  journal={bioRxiv},
  year={2026},
  doi={10.64898/2026.01.21.700911}
}