DeepCC
A pathway-informed deep learning framework for robust cancer molecular subtype classification.
Overview
DeepCC is an early project in the lab’s translational omics line, focused on making molecular subtyping more robust and clinically usable. Instead of classifying patients directly from fragile gene signatures, the framework converts expression data into biologically meaningful functional spectra and performs classification with a trainable neural network.
This pathway-informed design improves subtype separation, reduces the number of unclassifiable samples, and supports single-sample prediction across platforms. DeepCC established several of the robustness principles that later reappear in newer stratification systems.