Integrated multi-omics and machine learning reveal a gefitinib resistance signature for prognosis and treatment response in lung adenocarcinoma

Abstract

Gefitinib resistance (GR) presents a significant challenge in treating lung adenocarcinoma (LUAD), highlighting the need for alternative therapies. This study explores the genetic basis of GR to improve prediction, prevention, and treatment strategies. We utilized public databases to obtain GR gene sets, single-cell data, and transcriptome data, applying univariate and multivariate regression analyses alongside machine learning to identify key genes and develop a predictive signature. The signature's performance was evaluated using survival analysis and time-dependent ROC curves on internal and external datasets. Enrichment and tumor immune microenvironment analyses were conducted to understand the mechanistic roles of the signature genes in GR. Our analysis identified a robust 22-gene signature with strong predictive performance across validation datasets. This signature was significantly associated with chromosomal processes, DNA replication, immune cell infiltration, and various immune scores based on enrichment and tumor microenvironment analyses. Importantly, the signature also showed potential in predicting the efficacy of immunotherapy in LUAD patients. Moreover, we identified alternative agents to gefitinib that could offer improved therapeutic outcomes for high-risk and low-risk patient groups, thereby guiding treatment strategies for gefitinib-resistant patients. In conclusion, the 22-gene signature not only predicts prognosis and immunotherapy efficacy in gefitinib-resistant LUAD patients but also provides novel insights into non-immunotherapy treatment options.