| Description |
Cancer detection using plasma cell-free DNA (cfDNA) fragmentomics is an emerging research area, but little is known in pancreatic and biliary tract cancers. The aim of this study is to characterize the cfDNA fragmentomics in biliopancreatic cancers and develop an accurate method for cancer detection. A total of 218 individuals, including 71 non-cancer volunteers and 147 patients with biliopancreatic cancers (55 with cholangiocarcinoma (CCA), 30 with gallbladder cancer (GBC), 62 with pancreatic cancer (PAC)) were included in this study, and were then divided into a training cohort (16 CCAs, 14 GBCs, 28 PACs, and 31 non-cancer volunteers) and a validation cohort (39 CCAs, 16 GBCs, 34 PACs, and 40 non-cancer volunteers). Plasma cfDNA samples were obtained and subjected to low-coverage whole-genome sequencing (WGS) with a median coverage of 2.9x. Three cfDNA fragmentomic features including fragment size, end motif and nucleosome footprint were subjected to construct a stacked machine learning model for cancer detection. The stacked model presented robust performance for cancer detection in the validation cohort (area under curve (AUC): 0.941) and remained consistent even when using extremely low-coverage sequencing depth of 0.5x (AUC: 0.905). Further analysis showed that cfDNA fragmentomics served as an important diagnostic method in carbohydrate antigen 19-9 (CA19-9) negative biliopancreatic cancers, and final combined model of cfDNA fragmentomics and CA19-9 exhibited very high accuracy in detection of biliopancreatic cancers (AUC: 0.995). Besides, our method could also be used to distinguish biliopancreatic cancer subtypes. In summary, our model demonstrated ultrasensitivity and provided an affordable way for accurate noninvasive biliopancreatic cancer screening in clinical practice. |
