Abstract
Evaluating the accuracy of pulmonary nodule diagnosis avoids repeated low-dose computed tomography (LDCT)/CT scans or invasive examination, yet remains a main clinical challenge. Screening for new diagnostic tools is urgent. Herein, we established a nomogram based on the diagnostic signature of five circulating tsRNAs and CT information to predict malignant pulmonary nodules. In total, 249 blood samples of patients with pulmonary nodules were selected from three different lung cancer centers. Five tsRNAs were identified in the discovery and training cohorts and the diagnostic signature was established by the randomForest algorithm (tRF-Ser-TGA-003, tRF-Val-CAC-005, tRF-Ala-AGC-060, tRF-Val-CAC-024, and tiRNA-Gln-TTG-001). A nomogram was developed by combining tsRNA signature and CT information. The high level of accuracy was identified in an internal validation cohort (n = 83, area under the receiver operating characteristic curve [AUC] = 0.930, sensitivity 100.0%, specificity 73.8%) and external validation cohort (n = 66, AUC = 0.943, sensitivity 100.0%, specificity 86.8%). Furthermore, the diagnostic ability of our model discriminating invasive malignant ones from noninvasive lesions was assessed. A robust performance was achieved in the diagnosis of invasive malignant lesions in both training and validation cohorts (discovery cohort: AUC = 0.850, sensitivity 86.0%, specificity 81.4%; internal validation cohort: AUC = 0.784, sensitivity 78.8%, specificity 78.1%; and external validation cohort: AUC = 0.837, sensitivity 85.7%, specificity 84.0%). This novel circulating tsRNA-based diagnostic model has potential significance in predicting malignant pulmonary nodules. Application of the model could improve the accuracy of pulmonary nodule diagnosis and optimize surgical plans.