The paper introduces Cancer-Finder, a domain generalization-based deep-learning algorithm designed to accurately identify malignant cells in single-cell and spatial transcriptomics data. Single-cell and spatial transcriptomics are increasingly used to study cancer, but current algorithms lack accuracy and generalization, making it challenging to consistently identify malignant cells. Cancer-Finder addresses this issue by learning a generalization model from multiple datasets with varying distributions, enabling it to distinguish malignant and normal cells in single-cell data with an average accuracy of 95.16%. By replacing single-cell training data with spatial transcriptomic datasets, Cancer-Finder can also accurately identify malignant spots on spatial slides. The algorithm was applied to five clear cell renal cell carcinoma (ccRCC) spatial transcriptomic samples, identifying a gene signature of 10 genes that are significantly co-localized and enriched at the tumor-normal interface, with strong correlation to patient prognosis. Cancer-Finder is an efficient and extensible tool for malignant cell annotation, facilitating the discovery of biological mechanisms using single-cell and spatial transcriptomics data.The paper introduces Cancer-Finder, a domain generalization-based deep-learning algorithm designed to accurately identify malignant cells in single-cell and spatial transcriptomics data. Single-cell and spatial transcriptomics are increasingly used to study cancer, but current algorithms lack accuracy and generalization, making it challenging to consistently identify malignant cells. Cancer-Finder addresses this issue by learning a generalization model from multiple datasets with varying distributions, enabling it to distinguish malignant and normal cells in single-cell data with an average accuracy of 95.16%. By replacing single-cell training data with spatial transcriptomic datasets, Cancer-Finder can also accurately identify malignant spots on spatial slides. The algorithm was applied to five clear cell renal cell carcinoma (ccRCC) spatial transcriptomic samples, identifying a gene signature of 10 genes that are significantly co-localized and enriched at the tumor-normal interface, with strong correlation to patient prognosis. Cancer-Finder is an efficient and extensible tool for malignant cell annotation, facilitating the discovery of biological mechanisms using single-cell and spatial transcriptomics data.