No abstract

No abstract

No abstract

No abstract

Marker screening was performed on papillary thyroid cancer (PTC, 37 tissues, 55 plasmas), benign nodules (37 tissues, 55 plasmas) and normal samples (20 buffy coats; 123 plasmas) by MONOD+ assay. We identified differential markers by Wilcox rank sum test using Benjamini-Hochberg procedure to control false discovery rate. Predication models were built using machine-learning algorithms including random forest and support vector machine. They were validated using public DNA methylation data of thyroid tissues. Candidate markers were developed into a targeted sequencing panel and were validated on plasma DNA samples (115 PTC, 102 benign nodules). Best-performing markers were developed into an improved panel to classify additional plasma DNA samples of malignant or benign thyroid nodules.Results: From the MONOD+ data we identified over 1000 DNA methylation markers significantly differential between malignant and benign nodules. We built a classification model by random forest method, which classified DNA methylation profiles of thyroid nodules at a sensitivity of 90.5% and a specificity of 91.9% (95% CI, 0.91-1.0). We produced a targeted sequencing panel using those markers and sequenced plasma DNA of PTC and benign nodules. Two thirds of them were used as a training cohort to build a prediction model, which classified the remaining samples at an accuracy of 72%. We selected the best-performing markers to build an advanced version of panel, which classified additional 300 plasma DNA samples of thyroid nodules with increased sequencing depth to improve the accuracy and consistency in classification.Conclusions: Our study demonstrates that DNA methylation markers can robustly differentiate thyroid nodules based on their malignancy. They are thus promising candidates to develop non-invasive diagnostics for thyroid cancer screening.