CanCellCap: Robust Cancer Cell Capture Across Tissue Types on Single-Cell RNA-Seq Data by Multi-Domain Learning

CanCellCap is a multi-domain learning framework designed to identify cancer cells from single-cell RNA sequencing (scRNA-seq) data, suitable for all tissues, cancers, and sequencing platforms. By integrating domain adversarial learning and Mixture of Experts (MoE), CanCellCap extracts both tissue-common and tissue-specific patterns from gene expression profiles. Its masking-reconstruction strategy addresses sequencing platform variability, ensuring robust performance across diverse datasets.