Non-invasive detection of biliary tract cancer by low-coverage whole genome sequencing from plasma cell-free DNA: A prospective cohort study

Wang, Xiang; Fu, Xiao-Hui; Qian, Zi-Liang; Zhao, Teng; Duan, An-Qi; Ruan, Xiang; Zhu, Bin; Yin, Lei; Zhang, Yong-Jie; Yu, Wen-Long

BACKGROUND: The diagnosis of biliary tract cancer (BTC) is challenging in clinical practice. We performed a prospective study to evaluate the value of plasma copy number variation (CNV) assays in diagnosing BTC. METHODS: 47 treatment-naïve patients with suspicious biliary lesions were recruited. Plasma samples were collected at admission. Cell-free DNA was analyzed by low coverage whole genome sequencing, followed by CNV analyses via a customized bioinformatics workflow, namely the ultrasensitive chromosomal aneuploidy detector. RESULTS: 29 patients were pathologically diagnosed as BTC, including 8 gallbladder cancers (GBCs) and 21 cholangiocarcinomas (CCs). Cancer patients had more CNV signals as compared with benign patients (26/29 vs. 2/18, P < 0.001). The most frequent copy number gains were chr3q (7/29) and chr8q (6/29). The most frequent copy number losses were chr7p (6/29), chr17p (6/29), and chr19p (6/29). The sensitivity and specificity of plasma CNV assays in diagnosing BTC were 89.7% and 88.9%, respectively. For CA 19-9 (cutoff: 37 U/ml), the sensitivity was 58.6% and the specificity was 72.2%. The diagnostic accuracy of CNV assays significantly outperformed CA 19-9 (AUC 0.91 vs. 0.62, P = 0.004). Compared with CA 19-9 alone, the adding of CNV profiles to CA 19-9 increased the sensitivity in diagnosing GBC (75.0% vs. 25.0%) and CC (100% vs. 52.4%). Higher CNV burden was also associated with decreased overall survival (Hazard ratio = 4.32, 95% CI 2.06–9.08, P = 0.033). DISCUSSION: Our results suggest that BTC harbors rich plasma CNV signals, and their assays might be useful for diagnosing BTC.