Accurate and early diagnosis of non-small cell lung cancer using aberrant fragmentomic features of circulating cell-free mitochondrial DNA.

Abstract

Early differential diagnosis of benign and malignant lung diseases (DBM) is a critical challenge in the clinical oncology of non-small cell lung cancer (NSCLC). We aim to develop a novel strategy utilizing circulating cell-free mtDNA (ccf-mtDNA) fragmentomics for accurate and early diagnosis of NSCLC. We analyzed capture-based ccf-mtDNA sequencing data of 2306 plasma samples from 1357 NSCLC patients, 432 benign lung diseases (BLD) patients, and 517 healthy controls (HCs) obtained from 3 hospitals. Subsequently, using ccf-mtDNA fragmentomic features, we developed models for DBM and Early Diagnosis of NSCLC (EDL). Our analysis revealed significantly aberrant fragmentomic features of ccf-mtDNA in NSCLC patients compared with those in BLD patients and HCs. Remarkably, the DBM model demonstrated remarkable capability to distinguish NSCLC from BLD, outperforming serum biomarkers with an AUC exceeding 0.9551 in 3 validation cohorts. Still, the DBM model exhibited superior diagnostic performance even for small nodules (<1 cm), achieving an AUC of 0.9151. Moreover, the DBM model demonstrated precise clinical management ability of pulmonary lesions, thereby avoiding unnecessary invasive procedures in BLD patients and preventing delayed treatment in NSCLC patients. Furthermore, the EDL model demonstrated outstanding performance in detecting stage 0-I NSCLC, with an AUC exceeding 0.9759. Our multicenter study provides a novel non-invasive approach using ccf-mtDNA fragmentomics for the DBM and early diagnosis of NSCLC, with potential applications in clinical decision-making in the management of NSCLC.