Overview of scSHEFT. The input to scSHEFT consists of gene expression count data, peak count data, and GAS. scSHEFT employs a peak-specific encoder to embed the peak count data and a shared gene-specific encoder for embedding the gene expression and GAS data. Two alignment strategies, intraomics and interomics, are introduced to optimize the cellular embeddings in the latent space, ensuring that similar cells are clustered both within and across different omics layers.