Prognostic lncRNA, miRNA, and mRNA Signatures in Papillary Thyroid Carcinoma

Abstract

The current focus in the treatment of papillary thyroid carcinoma (PTC) is tumor progression. The aim of this study was to build RNA-based classifiers and develop a comprehensive model to provide progression-free interval (PFI) risk prediction for PTC. The RNAseq data, miRNAseq data, and clinical information of PTC were downloaded from The Cancer Genome Atlas database. Based on the differently expressed RNAs, the least absolute shrinkage and selection operator (LASSO) Cox regression model was utilized to build the RNA-based classifiers for PFI of the patients with PTC. A 6-messenger RNA (mRNA)-based classifier, a 5-long non-coding RNA (lncRNA)-based classifier, and a 4-microRNA (miRNA)-based classifier were constructed to predict the PFI. Patients with high risk based on the constructed RNA-based classifiers had worse prognosis in Kaplan–Meier curve analysis with log-rank test. The areas under the curves of the first, third, and fifth years in the training and testing set were 0.83, 0.82, and 0.82 and 0.67, 0.72, and 0.73 for the 6-mRNA-based classifier, respectively; 0.75, 0.84, and 0.85 and 0.71, 0.67, and 0.71 for the 5-lncRNA-based classifier, respectively; and 0.70, 0.77, and 0.79 and 0.74, 0.67, and 0.66 for the 4-miRNA-based classifier, respectively. The prediction capability of the three RNA-based classifiers was superior to the TNM stage system. Furthermore, a nomogram based on the verified independent prognostic factors was established for the prognostic prediction. The C-index and calibration plots indicated good predictive accuracy of the nomogram. In summary, the 6-mRNA-based classifier and 5-lncRNA-based classifier constructed in this study were independent prognostic factors for PTC.