[Identification of SORD as A Key Gene Mediating Osimertinib Primary Resistance 
in EGFR-Mutant Lung Adenocarcinoma via Machine Learning and Methylomics].

[BACKGROUND] Lung adenocarcinoma (LUAD) is the most common subtype of non-small cell lung cancer (NSCLC), and epidermal growth factor receptor (EGFR) mutation is its primary molecular driver event. Although the third-generation tyrosine kinase inhibitors (TKIs) Osimertinib has become the standard first-line therapy for such patients, the development of drug resistance severely limits long-term survival benefits. However, increasing evidence suggests that epigenetic remodeling is a crucial non-genetic mechanism contributing to resistance. Specifically, hypermethylation of DNA promoter regions can assist tumor cells in evading drug cytotoxicity by silencing key tumor suppressor genes or metabolic regulatory genes. This study aims to identify key genes regulating Osimertinib sensitivity through machine learning and high-throughput screening, and to dissect the role of methylation modifications in regulating primary resistance to Osimertinib.

[METHODS] Based on The Cancer Genome Atlas (TCGA)-LUAD EGFR mutation cohort, the oncoPredict model was applied to predict Osimertinib half inhibitory concentration (IC50) values. Through multi-omics association analysis, differentially expressed genes in the transcriptome and methylome significantly associated with IC50 were screened, respectively. The intersection of the two omics datasets was identified based on the "hypermethylation-low expression" negative correlation pattern, and LASSO regression with Bootstrap validation was applied to screen for the most robust core genes. Finally, in EGFR-mutant LUAD cell lines, the effects of the core gene on Osimertinib resistance, cell proliferation, and apoptosis were validated via drug sensitivity assays, demethylation treatment, and gene overexpression/knockdown experiments.

[RESULTS] Multi-omics analysis and machine learning algorithms identified SORD as the core candidate gene, with the methylation level of its key site (cg06424894) showing a significant negative correlation with mRNA expression. In vitro experiments revealed that the Osimertinib-resistant cell line H1650 was characterized by SORD hypermethylation and low expression, and demethylation treatment significantly restored its expression; whereas sensitive strains showed the opposite trend. Functional experiments confirmed that overexpression of SORD in H1650 resistant cells could reverse resistance, inhibit cell proliferation, and promote apoptosis. Conversely, knockdown of SORD in sensitive H1975 and PC9 cells significantly induced resistance, promoted proliferation, and inhibited apoptosis.

[CONCLUSIONS] The methylation status or expression level of SORD holds promise as a potential biomarker for predicting Osimertinib efficacy, and targeting SORD-related pathways may provide new strategies for overcoming Osimertinib resistance.