A scRNA-seq Based Prediction Model of EGFR-TKIs Resistance in Patients With Non-Small Cell Lung Adenocarcinoma

Abstract

Background EGFR-TKIs were used in NSCLC LUAD patients with actionable EGFR mutations and prolonged the overall survival. However, most patients treated with EGFR-TKIs developed resistance within a median of 10 to 14 months. EGFR-TKIs resistance risk prediction will help individualized management of patients with potential risk. Method We built an R-index model trained by single-cell RNA (scRNA) data with the OCLR algorithm. We then validated the accuracy of the model in multiple datasets and evaluated the performance with orthogonal verification by scRNA data and three large cohorts data in the aspects of EGFR-TKIs resistance pathways and immune microenvironment. Results When applying the R-index model in cell lines, mouse xenograft models, and three large LUAD cohorts(n=892) to perform verification analysis, we found that the R-index was significantly related to the dynamic changes of cell numbers, the osimertinib resistance status of mice, and the outcome of the cohort. We also found that the glycolysis pathway and the KRAS up-regulation pathway were related to EGFR-TKIs resistance. And MDSC was a major factor of immunosuppression in the resistant microenvironment. Conclusions Through in vivo and in vitro validation, the R-index model based on scRNA sequencing data was confirmed containing the capability of predicting the EGFR-TKIs resistance. We also used scRNA data and cohort data to orthogonally verify the performance of the R-index. These results suggested that R-index could be used as an indicator of EGFR-TKIs resistance prediction in preclinical studies.

Keywords: EGFR-TKI resistance, lung adenocarcinoma, scRNA-seq, R-index