Integrating Machine Learning to Identify Prognostic Subtypes and the Cisplatin Resistance Gene RRM2 in Lung Adenocarcinoma

Lung adenocarcinoma (LUAD) is a highly invasive malignancy associated with poor prognosis. However, reliable molecular subtypes and prognostic signatures for LUAD, particularly their associations with tumor immunity and treatment response, remain poorly characterized. This study aimed to identify distinct molecular subtypes and establish a robust prognostic signature for LUAD, while exploring potential therapeutic targets associated with cisplatin resistance. In this study, we integrated ten clustering algorithms with 101 machine learning methods to identify two distinct molecular subtypes of LUAD and established a robust consensus machine learning-derived signature (CMLS). Patients belonging to the Cluster A subtype exhibited a more favorable prognosis and an enhanced immune response. Furthermore, a CMLS model constructed from 19 genes exhibited strong predictive power. Individuals with high CMLS scores displayed reduced CD4 + T-cell infiltration, poorer survival outcomes, and diminished responses to immunotherapy, whereas those with the low CMLS scores showed improved clinical outcomes, better immunotherapy responsiveness, and features of an immunologically "hot" tumor phenotype. Among the identified genes, ribonucleotide reductase subunit M2 (RRM2) emerged as a core regulator associated with LUAD prognosis. RRM2 was markedly upregulated in lung cancer tissues and the LUAD cell lines and was associated with the cisplatin (DDP) resistance. Suppression of RRM2 expression significantly enhanced the cisplatin sensitivity of A549/DDP cells. In summary, our findings establish a molecular classifier that stratifies LUAD patients into distinct risk categories and highlight RRM2 as a promising biomarker and therapeutic target associated with cisplatin resistance.