Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

Poster Display session 2

5051 - Classification of thyroid nodule using DNA methylation profiling on tissue and circulating tumor DNA

Date

29 Sep 2019

Session

Poster Display session 2

Presenters

Shubin Hong

Citation

Annals of Oncology (2019) 30 (suppl_5): v756-v759. 10.1093/annonc/mdz267

Authors

S. Hong1, J. Li2, L. Cheng3, S. Yu1, Z. Zhang3, B. Lin2, Z. Su3, Z. Ke4, R. Liu3, S. Peng5, Q. Li6, Q. Zhang6, Z. Guo6, W. Lv2, H. Xiao1

Author affiliations

  • 1 Endocrinology, 1st Affiliated Hospital of Sun Yat-sen University, 510080 - Guangzhou/CN
  • 2 Thyroid And Breast Surgery, 1st Affiliated Hospital of Sun Yat-sen University, 510080 - Guangzhou/CN
  • 3 Research & Development, Singlera Genomics Inc. and Singlera Genomics (Shanghai) Ltd., 201203 - Shanghai/CN
  • 4 Pathology, 1st Affiliated Hospital of Sun Yat-sen University, 510080 - Guangzhou/CN
  • 5 Clinical Trials Unit, 1st Affiliated Hospital of Sun Yat-sen University, 510080 - Guangzhou/CN
  • 6 Head And Neck Surgery, Sun Yat-sen University Cancer Center, 510060 - Guangzhou/CN
More

Resources

Abstract 5051

Background

A rapid increase in incidence of thyroid cancer has been reported in recent years. However, ambiguities between benign and malignant nodules in cytological features have led to unnecessary thyroidectomy or over-treatment. We aimed to identify DNA methylation markers to accurately classify thyroid nodules, which may be developed into non-invasive screening diagnostics for thyroid cancer to reduce unnecessary thyroid removals.

Methods

Genome-wide DNA methylation profiles of papillary thyroid cancer nodules (tissue, N = 37; plasma, N = 55), benign nodules (tissue, N = 37; plasma, N = 55) and healthy samples (buffy coat, N = 20; plasma, N = 123) were generated by a modified RRBS method. An independent RRBS dataset including 89 malignant and 72 benign nodules was derived from Yim et al. 2019. We adopted Singlera’s MONOD+ assay to interrogate the methylation state of over 4,000,000 CpG sites across more than 200,000 methylation haplotype blocks (MHBs). We identified discriminatory DNA methylation MHBs by Wilcox rank sum test. Using our dataset as a training set, we built several diagnostic models to classify thyroid nodules using machine learning methods. We validated the model by classifying public methylation datasets of thyroid tissues.

Results

We generated an initial list of top 1000 DNA methylation markers and built a classification model by the random forest method. We classified DNA methylation profiles of thyroid nodules published by Yim et al. 2019. Results show that this model achieved a sensitivity of 82%, a specificity of 94.4% and AUC of 0.94 (95% CI, 0.91-0.98). We randomly selected the 2/3 plasma samples as training data to build a prediction model. The model yielded an accuracy of 72% in the validation cohort. We also identified 2392 deferentially methylated regions by comparing benign tissues with malignant tissues. DMR associated genes are highly enriched in many cancer related pathways and MSigDB immunologic signatures, indicating the development of thyroid cancer is closely associated with immune dysfunction.

Conclusions

Our study demonstrates that DNA methylation markers can robustly differentiate thyroid nodules. They are thus promising candidates to develop non-invasive diagnostics for thyroid cancer screening.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The First Affiliated Hospital of Sun Yat-sen University.

Funding

National Natural Science Foundation of China (81772850).

Disclosure

All authors have declared no conflicts of interest.

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings