Prediction of tissue-of-origin of early-stage cancers using serum miRNomes

Large-scale serum miRNomics in combination with machine learning could lead to the development of a blood-based cancer classification system. Serum microRNA profiles of 9,921 patients with 13 types of human solid cancers [breast (BR; n = 675), bladder (BL; n = 399), biliary tract (BT; n = 402), colorectal (CR; n = 1,596), esophageal squamous cell (ES; n = 566), lung (LU; n = 1,699), gastric (GA; n = 1,418), hepatocellular (HC; n = 348), pancreatic (PA; n = 851), prostate (PR; n = 1,027), ovarian (OV; n = 400), bone and soft tissue sarcomas (SA; n = 299), and intraparenchymal brain tumors (gliomas, primary central nervous system lymphomas, and metastatic tumors) (GL; n = 241)] ,626 non-cancer conditions [benign disease in the breast (BR_N; n = 31), extraparenchymal brain tumor and benign disease in the brain (GL_N; n = 24), benign disease in the ovary (OV_N; n = 28), benign disease in the prostate (PR_N; n = 230), or benign disease in the bone and soft tissue (SA_N; n = 313)], and 5,643 non-cancer controls.