CancerSubtyper: a deep learning framework for cancer subtyping through DNA methylation data

Molecular subtyping is essential for precision oncology, enabling the classification of tumors into biologically and clinically relevant categories. DNA methylation has emerged as a promising biomarker for cancer subtyping, yet its application remains limited by high dimensionality, batch effects, and the lack of automated, user-friendly analytical tools.

Here, we present CancerSubtyper, an end-to-end computational framework for deep learning-based cancer subtyping using DNA methylation data, which is accessible through an intuitive web interface designed to support interactive exploration and downstream analysis. CancerSubtyper integrates two complementary models: a semi-supervised classifier for cancers with well-established subtypes, and a hybrid framework that integrates supervised and unsupervised learning to identify novel subtypes. The framework automatically performs preprocessing, feature selection, batch correction, and cancer subtyping while offering interactive visualization for subtype exploration and validation.

By providing an automated, end-to-end workflow accessible through a user-friendly web interface, CancerSubtyper lowers the barrier to large-scale methylation analysis and provides a powerful tool for molecular subtyping and precision oncology research. The framework is freely accessible at https://github.com/ycheung5/cancersubtyper/ .