Abstract
BACKGROUND: Anoikis and epithelial-mesenchymal transition (EMT) are pivotal in the distant metastasis of lung adenocarcinoma (LUAD). A detailed understanding of their interplay and the identification of key genes is vital for effective therapeutic strategies against LUAD metastasis. METHODS: Key prognostic genes related to anoikis and EMT were identified through univariate Cox regression analysis. We utilized ten machine learning algorithms to develop the Anoikis and EMT-Related Optimal Model (AEOM). The TCGA-LUAD dataset served as the training cohort, while six additional international multicenter LUAD datasets were employed as validation cohorts. The average concordance index (c-index) was used to evaluate model performance and identify the most effective model. Subsequent multi-omics analyses were conducted to explore differences in pathway enrichment, immune infiltration, and mutation landscapes between high and low AEOM groups. Experimental validation demonstrated that RHPN2, a key biomarker within the model, acts as an oncogene facilitating LUAD progression. RESULTS: The AEOM displayed superior prognostic predictive performance for LUAD patients, outperforming numerous previously published LUAD signatures. Biologically, the AEOM was notably associated with immune features; the high AEOM group exhibited decreased immune activity and a tendency towards immune-cold tumors, as well as a higher tumor mutational burden (TMB). Subgroup analysis revealed that the low AEOM + high TMB group had the most favorable prognosis. The high AEOM group was primarily enriched in cell cycle-related pathways, promoting cancer cell proliferation. RHPN2, a crucial gene within the AEOM (correlation = 0.85, P < 0.05), was linked to poorer prognosis in LUAD patients with elevated RHPN2 expression. Further in vitro experiments showed that RHPN2 modulates LUAD cell proliferation and invasion. CONCLUSION: The AEOM provides a robust prognostic model for LUAD, uncovering critical immune and biological pathways, with RHPN2 identified as a key oncogenic driver. These findings offer valuable insights for targeted therapies and enhanced patient outcomes.