Machine learning reveals two heterogeneous subtypes to assist immune therapy based on lipid metabolism in lung adenocarcinoma

BackgroundLipid metabolism pivotally contributes to the incidence and development of lung adenocarcinoma (LUAD). The interaction of lipid metabolism and tumor microenvironment (TME) has become a new research direction. MethodsUsing the 1107 LUAD records from the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, a comprehensive exploration was performed on the heterogeneous lipid metabolism subtypes based on lipid metabolism genes (LMGs) and immune-related genes (LRGs). The clinical significance, functional status, TME interaction and genomic changes of different subtypes were further studied. A new scoring system, lipid-immune score (LIS), was developed and validated. ResultsTwo heterogeneous subtypes, which express more LMGs and show the characteristics of tumor metabolism and proliferation, are defined as lipid metabolism phenotypes. The prognosis of lipid metabolism phenotype is poor, and it is more common in patients with tumor progression. Expressing more IRGs, enrichment of immunoactive pathways and infiltration of effector immune cells are defined as immunoactive phenotypes. The immunoactive phenotype has a better prognosis and stronger anti-tumor immunity and is more sensitive to immunotherapy. In addition, KEAP1 is a driving mutant gene in the lipid metabolism subtype. Finally, LIS was developed and confirmed to be a robust predictor of overall survival (OS) and immunotherapy in LUAD patients. ConclusionTwo heterogeneous subtypes of LUAD (lipid metabolism subtype and immune activity subtype) were identified to evaluate prognosis and immunotherapy sensitivity. Our research promotes the understanding of the interaction between lipid metabolism and TME and offers a novel direction for clinical management and precision therapy aimed to LUAD patients.