Machine learning prognostic model and drug survival analysis for lung adenocarcinoma in the context of radiotherapy

Abstract: Background: Patients with lung adenocarcinoma (LUAD) receiving radiotherapy represent an important but underexplored clinical subgroup. These patients often undergo concomitant pharmacologic treatments, yet the prognostic impact and underlying determinants of such combined regimens remain poorly understood. Objective: This retrospective observational study aimed to develop and validate a radiotherapy-specific machine learning prognostic model for LUAD and to compare survival across concomitant pharmacologic regimens. Methods: In this retrospective observational study, using genomic and clinical data from TCGA, a radiotherapy-specific prognostic model for LUAD was developed and validated through ten machine learning algorithms. Survival analyses were conducted across distinct concomitant pharmacologic strategies, followed by functional enrichment to elucidate molecular mechanisms underlying differential outcomes. Results: Demonstrating robust prognostic abilities, the model efficiently sorted patients into high- and low-risk categories. Both treatment type and risk score independently predicted overall survival, with significant interaction effects. Low-risk patients receiving targeted or combination therapy—mainly erlotinib, gefitinib, or bevacizumab—exhibited substantially improved survival compared with those receiving conventional chemotherapy. Enrichment of "Exogenous peptide presentation," "MHC class II assembly," "Peptide–MHC II assembly," and "Symbiotic interaction" pathways indicated immune modulation and host–tumor crosstalk as key mediators of treatment efficacy. Conclusion: This study establishes a radiotherapy-specific prognostic model for lung adenocarcinoma, demonstrating distinct molecular and therapeutic heterogeneity and highlighting the superior survival benefit of targeted combination therapy in low-risk patients.