Abstract
The differentiation between benign and malignant pulmonary ground-glass nodules (GGNs) has always been a current clinical hotspot issue. Patients with pulmonary GGNs from 8 centers were enrolled to identify biomarkers for malignancy and benignity discrimination, and an integrated biomarker panel comprising two miRNA, one long non-coding RNA (lncRNA), and one circular RNA (cirRNA) identified by multivariate logistic regression analysis were established. The classifier had area under the receiver-operating characteristic (ROC) curve (AUC) of 0.88, 87.8% sensitivity and 70% specificity, being significantly higher compared with the Mayo model (AUC of 0.72), Brock model (AUC of 0.70), and Herder model (AUC of 0.82) (all p < 0.05). Moreover, relative expression of miR1246 and miR-122 gradually increased from adenocarcinoma in situ (AIS) to minimally invasive adenocarcinoma (MIA) to invasive adenocarcinoma (IAC). The classifier was validated in two independent sets of patients. It has been proved that the integration of three kinds of non-coding RNAs (ncRNAs) could more accurately identify early staged lung cancer among indeterminate GGNs.