Identification of key gene signatures for predicting chemo-immunotherapy efficacy in extensive-stage small-cell lung cancer using machine learning

Objectives: The lack of definitive biomarkers presents a significant challenge for chemo-immunotherapy in extensive-stage small-cell lung cancer (ES-SCLC). We aimed to identify key genes associated with chemoimmunotherapy efficacy in ES-SCLC through comprehensive gene expression analysis using machine learning (ML). Methods: A prospective multicenter cohort of patients with ES-SCLC who received first-line chemo-immunotherapy was analyzed. RNA sequencing was performed on tumor samples to assess gene expression levels. ML techniques were applied to identify key gene features associated with treatment efficacy. A panel of genes was then developed and validated using the nCounter system, and the model's performance in predicting 180-day progression-free survival (PFS) was evaluated. Results: A total of 93 patients were included in the analysis. ML-based gene selection identified a gene set comprising 83 genes from the comprehensive gene expression data. An nCounter panel was developed using these genes, and an ML model was developed based on their expression levels. In the validation set, the model achieved an accuracy of 0.93, precision of 1.00, a true positive rate of 0.83, and a true negative rate of 1.00. PFS was significantly longer in the high-efficacy group than in the low-efficacy group in the validation set (P < 0.001). Conclusions: These findings provide a foundation for biomarker development in ES-SCLC and highlight the potential of this method as a cost-effective, simple, and rapid tool for assessing treatment efficacy in clinical practice.