The aim of this study was to form a novel and rational diagnostic model for CCA with several biomarkers expressed in bile and blood by machine learning. In this study several potential protein biomarkers were screened by iTRAQ-based MS analysis from 8 bile specimens from CCA and bile duct stones patients respectively.

However, the content of bile is complex, and other reasons such as liver or inflammatory processes can also influence the expression of some secreted proteins. And the most valuable cancer protein markers are the ones produced and secreted by cancer cells, so in order to accurately observe the proteins secreted by CCA cells, supernatants of four CCA cell lines and human intrahepatic biliary epithelial cells (HIBEpic) were collected for label-free and 6 pairs of CCA tissues were used for iTRAQ-based MS analysis.

Finally, two proteins were differently expressed all in bile, supernatant and tissues, the one overexpressed in cancer group was selected for further study. Previous studies have shown that the diagnostic value of a biomarker can be improved when combining other circulating biomarkers with machine learning. Fortunately, we have found that the level of several biomarkers in routine blood tests would change in CCA process, so we attempted to establish an efficient CCA diagnostic model based on the algorithmic combination of random forest (RF) method and multi bile protein biomarkers and serum biomarkers to achieve CCA control and management. Ultimately, we aimed to develop a multi-analyte panel with machine learning to efficiently distinguish CCA from bile duct stones.