M2OST: Many-to-one Regression for Predicting Spatial Transcriptomics from Digital Pathology Images

This is the official repository for AAAI 2025 paper M2OST: Many-to-one Regression for Predicting Spatial Transcriptomics from Digital Pathology Images

The code of the core M2OST network will be released shortly after in a torch.nn.Module style, while the complete training/validation code will be updated after a comprehensive sweep. Thank you for your interest in our work.

Update 2025/04/16

Sorry for the late release. As I have finished my diseertatioon today, all the related codes are going to be released very soon. Right now a roughly updated version have been pushed, but please note that the pretrained weights are not verified yet and they may be from my previous project (see M2ORT). I will soon check all the weights and only keep the ones that can fit M2OST.

To run the code, please follow the instructions below:

1. Clone the code

Clone the code by running:

git clone git@github.com/Dootmaan/M2OST.git

then cd into this directory.

2. Prepare your datasets

Download the datasets from their official site.

• HBC: https://data.mendeley.com/datasets/29ntw7sh4r/5.
  • Make sure you have also downloaded this file before using the HBC dataset.
• HER2+: https://zenodo.org/records/3957257#.Y4LB-rLMIfg.
  • The files are encrypted by 7z. To decrypt these files, use the following passwords:
    • count matrices and images: zNLXkYk3Q9znUseS
    • meta data and spot selection: yUx44SzG6NdB32gY
• cSCC: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE144240

3. Train or validate the model

Start training the M2ORT model using the following command:

CUDA_VISIBLE_DEVICES=0,1 nohup python3 -u train_m2ost_example.py >train_m2ost_example.log 2>&1

Please note that the train_m2ost_example.py is for HBC dataset. You will have to modify the code accordingly when testing on the HER2+, cSCC or other datasets. You can also refer to the train_m2ort.py file in the M2ORT repo.