StHistDDComp

# StHistDDComp ### _Spatial Transcriptomics Histology-based Domain Detection Comparison_ --- ## Overview Currently, the repository supports and integrates the following methods: 1. [stLearn](https://github.com/BiomedicalMachineLearning/stLearn)2. [SpaGCN](https://github.com/jianhuupenn/SpaGCN)3. [conST](https://github.com/ys-zong/conST)4. [DeepST](https://github.com/JiangBioLab/DeepST)5. [ScribbleDom](https://github.com/1alnoman/ScribbleDom) Each method is wrapped with a uniform interface for running experiments, saving results, and ensuring reproducibility across random seeds. --- ## Repository Structure ```txt StHistDDComp/ │ ├── README.md ├── conST_run/ # Runner and setup scripts for conST │ ├── README.md │ ├── run_20.py │ └── setup_conST_run.sh ├── datasets/ # Dataset folder (DLPFC, HBC, etc.) │ └── ... ├── DeepST_run/ # Runner and setup scripts for DeepST │ ├── README.md │ ├── config.json │ ├── run_20.py │ └── setup_DeepST_run.sh ├── results/ # Output folder for domain labels and metrics ├── run_analysis/ # Analysis scripts for metrics and visualization │ ├── config.json │ └── wilcox.py ├── ScribbleDom_run/ # Runner and setup scripts for ScribbleDom │ ├── README.md │ ├── config.json │ ├── run_20.py │ ├── ScribbleDom_run.py │ └── setup_ScribbleDom_run.sh ├── SpaGCN_run/ # Runner and setup scripts for SpaGCN │ ├── README.md │ ├── config.json │ ├── run_20.py │ └── setup_SpaGCN_run.sh └── stLearn_run/ # Runner and setup scripts for stLearn ├── README.md ├── config.json ├── run_20.py └── setup_stLearn_run.sh``` --- ## Setup Instructions ### 1\. Clone the repository ```bashgit clone https://github.com/asifajrof/StHistDDComp.gitcd StHistDDComp``` ### 2\. Prepare datasets Download and place your spatial transcriptomics datasets in the `datasets/` directory. By default, the pipeline expects: ```txt datasets/ └── DLPFC/ ├── 151507/ │ └── filtered_feature_bc_matrix.h5 ├── 151508/ └── ...``` --- ## Running the Methods Each method has its own environment and setup instructions. You can run them independently as described below. --- ### **conST** #### Setup ```bashcd conST_runconda create -n conST_run python=3.10conda activate conST_runbash setup_conST_run.sh``` #### MAE model weights download download the model weights from [here](https://drive.google.com/file/d/1I7uLoL8ay8Scu_sYdjPg1jRaOtcpTNnP/view?usp=drive_link) #### Configuration Edit `/conST_run/config.json` to customize. #### Run ```bashpython conST_run.py <path/to/config.json> <seed>``` Or run multiple seeds automatically: ```bashpython run_20.py DLPFC``` **Outputs:** - `results/<histology_use>/results_seed_<seed>/<sample_name>/labels.csv` Contains columns: - `label` – domain assignments - `refined_label` – post-refinement domains --- ### **DeepST** #### Setup ```bashcd DeepST_runconda create -n DeepST_run python=3.10conda activate DeepST_runbash setup_DeepST_run.sh``` #### Configuration Edit `/DeepST_run/config.json` to customize. #### Run ```bashpython DeepST_run.py <path/to/config.json> <seed>``` Or run multiple seeds automatically: ```bashpython run_20.py DLPFC``` **Outputs:** - `results/w_hist/results_seed_<seed>/<sample_name>/labels.csv` Contains columns: - `label` – domain assignments - `refined_label` – post-refinement domains --- ### **SpaGCN** #### Setup ```bashcd SpaGCN_runconda create -n SpaGCN_run python=3.10conda activate SpaGCN_runbash setup_SpaGCN_run.sh``` #### Configuration Edit `/SpaGCN_run/config.json` to customize. #### Run ```bashpython SpaGCN_run.py <path/to/config.json> <seed>``` Or run multiple seeds automatically: ```bashpython run_20.py DLPFC``` **Outputs:** - `results/<histology_use>/results_seed_<seed>/<sample_name>/labels.csv` Contains columns: - `label` – domain assignments - `refined_label` – post-refinement domains --- ### **stLearn** #### Setup ```bashcd stLearn_runconda create -n stLearn_run python=3.10conda activate stLearn_runbash setup_stLearn_run.sh``` #### Configuration Edit `/stLearn_run/config.json` to customize. #### Run ```bashpython stLearn_run.py <path/to/config.json> <seed>``` Or run multiple seeds automatically: ```bashpython run_20.py DLPFC``` **Outputs:** - `results/w_hist/results_seed_<seed>/<sample_name>/labels.csv` Contains columns: - `label` – domain assignments - `refined_label` – post-refinement domains --- ### **ScribbleDom** #### Setup ```bashcd ScribbleDom_rungit clone https://github.com/1alnoman/ScribbleDom.gitcd ScribbleDomconda env create -f environment.ymlcd ..``` #### Run ```bashconda activate scribble_dompython run_20.py``` This executes **ScribbleDom_run.py**, which: - Updates seed values dynamically- Runs the model on DLPFC and HBC datasets- Saves results in `ScribbleDom_run/results/<seed>/` --- ## Reproducibility Each runner script supports a **`seed`** parameter to ensure reproducibility. Example: ```bashpython DeepST_run.py config.json 123``` The framework also provides `run_20.py` scripts to batch-run experiments across 20 predefined seeds for robust comparison. --- Data - https://drive.google.com/drive/folders/1Dj2f8JlJGBpOxx1yEfNs0WWzrBq3hRKN?usp=sharingResults - https://drive.google.com/drive/folders/1bP-MZtCpAIwonRhDFwiFBbkTxIZgemGY?usp=sharing # Analysis For analysis from follow the guideline in Analysis/README.md. It will download the data from: (https://drive.google.com/drive/folders/1Dj2f8JlJGBpOxx1yEfNs0WWzrBq3hRKN?usp=sharing) and execute all the analysis if you don't have any /Analysis/data/ folder. If you want to run analysis on your own data, keep your data in, Analysis/Data folder. For getting the data accross all the method download the data from: https://drive.google.com/drive/folders/1Dj2f8JlJGBpOxx1yEfNs0WWzrBq3hRKN?usp=sharing For getting the reproducible result downlonload the results from: https://drive.google.com/drive/folders/1bP-MZtCpAIwonRhDFwiFBbkTxIZgemGY?usp=sharing