Integrative subtyping of nonsmall cell lung cancer using histopathology and multi-omics data

Abstract

Nonsmall cell lung cancer (NSCLC), encompassing lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), is a major global health challenge due to its high mortality rate. Current molecular classifications of NSCLC fail to adequately integrate subtype-specific molecular and phenotypic differences, and many are not directly applicable to clinical diagnosis, treatment, or prognosis guidance. To address this, we develop a machine learning-based tumor subtyping framework, Morphgene, that integrates morphological analysis from Hematoxylin and Eosin (H&E) stained slides with multiomics data, successfully delineating four distinct survival-related subtypes for both LUAD and LUSC. Our analysis identifies unique molecular profiles and treatment responses for these subtypes: LUAD’s Cluster C is characterized by low mutation rates and EGFR mutations, showing resistance to immunotherapy but sensitivity to targeted therapies. In contrast, LUAD’s Cluster B and LUSC’s Cluster D are likely to benefit from immunotherapy. LUSC’s Cluster A also shows enhanced survival with chemoradiotherapy. This integrated subtyping approach provides clearer insights for personalized treatment strategies in NSCLC.