CASPER: Cross-modal Alignment of Spatial and single-cell Profiles for Expression Recovery

This repository contains the implementation of CASPER, a cross-attention–based deep learning framework for predicting unmeasured gene expression in spatial transcriptomics (ST) using matched single-cell RNA-seq (scRNA-seq) data.
The model learns how spatial spots interact with transcriptional cell-type centroids through a dual-encoder attention mechanism, enabling accurate imputation of missing genes across multiple ST technologies.

Running the Code

Training

Run the following command to start training:

python main.py --config AttentionParams.json