CA-CAE Enables Pan-Cancer Subtype Classification And Prognosis

Researchers Zhang et al. (published February 20, 2026) introduce CA-CAE, a convolutional autoencoder with a channel-attention mechanism that integrates gene expression, DNA methylation, and microRNA data to identify survival-associated cancer subtypes and predict prognosis. Applied across 15 cancer types, CA-CAE demonstrates superior survival-prediction performance versus traditional statistical and other deep-learning methods, with code and data publicly available.