Abstract 3345: Tissue of origin (TOO) localization using targeted methylation biomarkers discovered by whole-genome bisulfite sequencing

Abstract

Abstract DNA methylation is a promising biomarker for tissue of origin (TOO) localization, which is important for classification of cancer of unknown primary site (CUP) in multiple cancer early detection. The current most of CpG sites of TOO focused on CpG islands based on The Cancer Genome Atlas (TCGA) data sets, but remains largely under-explored across the whole genome. In this study, we conducted the whole-genome bisulfite sequencing (WGBS) in 481 tumor samples and 361 matched adjacent normal tissue samples for 9 common types of cancers, including lung, breast, colon, liver, stomach, lymph, pancreas, esophageal and ovarian. After adjusted the tumor cells’ DNA methylation rates, 3,000 CpG sites, 60% of which were not on CpG islands, were selected through Random Forest classifier (multi-class) to identify cancer type. To evaluate these TOO methylation biomarkers, we used Random Forest Model to predict TOO of the methylation data sets in The Cancer Genome Atlas (TCGA), including 1,704 tumor samples. The results showed that accuracy of TOO could achieve 98.5% for these tumor samples of TCGA. Among these samples, the accuracy of TOO of breast, liver, lung, colorectal and pancreas were 99.6%, 99.9%, 97.7%, 99.2% and 99.3% respectively. Several samples from Esophagus and Stomach were classified to each other, and their accuracy of TOO was relatively low, 97.6% and 96.7%. It demonstrated that these methylation biomarkers could detect multiple cancers, and the multi-class model had good performance for TOO localization. Citation Format: Qiang Wei, Wanglong Deng, Zhihui Xu, Yihui Han, Yuan Yang, Chao Zhang, Xiaoqiang Wang, Xuan Tang, Chao Song, Qing Xu, Yong Ren. Tissue of origin (TOO) localization using targeted methylation biomarkers discovered by whole-genome bisulfite sequencing [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 3345.