Longitudinal single-cell RNA model aids prediction of EGFR-TKI resistance

Abstract Resistance is inevitable and a major challenge in treating lung adenocarcinoma (LUAD) patients with EGFR mutations. This study aimed to investigate the mechanism of EGFR-TKI resistance in LUAD using longitudinal single-cell RNA sequencing (scRNA-seq) data. We collected tumour samples of LUAD patients before and after EGFR inhibitor treatment and performed single-cell RNA sequencing. We used machine learning models for cell annotation and classified cells into subgroups. The inferCNV algorithm was used for CNV score calculation and tumour cell identification, and metabolic analysis was done using a gene-scoring approach. EGFR resistance score (ERscore), a gene signature derived from resistant tumour cells, was established to evaluate the predictiveness to EGFR-TKI inhibitors. The investigation classified subgroups of cells and identified three tumour cell types as critical cells mediating EGFR-TKI resistance. Our data also analysed the metabolic aspects of EGFR-TKI resistance using a single-cell approach. It showed that some tumour cell subtypes had a consistent metabolic profile, significantly up-regulating purine metabolism, oxidative phosphorylation, glycogen, and lipid metabolism. An assessment system called ERscore was established to evaluate the association between EGFR-TKI resistance and tumour ecosystem. The analysis showed a significant correlation between the ERscore and EGFR-TKI resistance, lung cancer phenotype, and prognosis. The findings suggest that the molecular mechanisms driving EGFR-TKI resistance in lung cancer may also contribute to poorer prognosis, particularly in lung adenocarcinomas with high EGFR mutation rates. Overall, the study provides important insights into the mechanisms of EGFR-TKI resistance in lung cancer at the single-cell level.