PREST

This is the official implementation for our paper:"Evidence-aware Integration and Domain Identification of Spatial Transcriptomics Data" (PREST: PRototype-based Evidence-aware integration framework for Spatial Transcriptomics data)

---

Overview

Experiments

Please replace the following {dataset_name} with the actual dataset name.

Setup

conda create -n PREST python=3.8.20
conda activate PREST
pip install -r requirements.txt

Data

To generate DLPFC data:

1. Set the datasets variable in generate_data_DLPFC.py to your target slice number (e.g., 151507)
2. Execute the generation script:

python generate_data_{dataset_name}.py

For All Other Datasets

Direct execution (no configuration needed):

python generate_data_{dataset_name}.py

Pretrain

python ./pretrain/ae/main.py --name {dataset_name} --lr 1e-3 --epoch 100
python ./pretrain/gae/main.py --name {dataset_name} --lr 1e-3 --epoch 100
python ./main.py --name {dataset_name} --lr 1e-3 --epoch 100 --mode ntrain

Train

python ./main.py --name {dataset_name} --epoch 200

Datasets

You can download the benchmark datasets from the links below:

• DorsoLateral PreFrontal Cortex (DLPFC): LieberInstitute/spatialLIBD: Code for the spatialLIBD R/Bioconductor package and shiny app
• Human Breast Cancer (HBC): Human Breast Cancer (Block A Section 1) - 10x Genomics
• Mouse Anterior Brain (MBA): Mouse Brain Serial Section 1 (Sagittal-Anterior) - 10x Genomics