Altered bile metabolome and its diagnostic potential for biliopancreatic malignancies

•Altered metabolic composition in bile from biliopancreatic malignancies can be used to differentiate malignant from benign disease.•The metabolic changes associated with biliopancreatic malignancies occurred mainly in lipid metabolites, among which fatty acid metabolism was most significantly altered.•A combined model for biliary tract cancer based on the fatty acid indexes in this study improved the diagnostic performance of existing serum tumor markers.•Different fatty acid metabolic profiles may contribute to the prognostic assessment of patients with biliopancreatic malignancies. Due to the difficulty of pathological sampling, the clinical differentiation between benign and malignant biliopancreatic diseases remains challenging. Endoscopic retrograde cholangiopancreatography (ERCP) is used to investigate biliary diseases, enabling the collection of bile. This study assessed potential metabolic alterations in biliopancreatic malignancies by exploring changes in the bile metabolome and the diagnostic potential of bile metabolome analysis. A total of 264 bile samples were collected from patients who were divided into a discovery cohort (n = 85) and a validation cohort (n = 179). Untargeted metabolomic analysis was used in the discovery cohort, while targeted metabolomic analysis was used in the validation cohort for further investigation of the differentially abundant metabolites. The untargeted metabolomic analysis revealed that the metabolic changes associated with biliopancreatic malignancies occurred mainly in lipid metabolites, among which fatty acid metabolism was most significantly altered, and differentially abundant metabolites identified in the discovery cohort were mainly enriched in unsaturated fatty acid synthesis and linolenic acid synthesis pathways. Analysis of free fatty acid (FFA) metabolism in the validation cohort revealed that the FFA levels and related indicators verified the abnormal fatty acid metabolism associated with biliopancreatic malignancies. The combined model for biliopancreatic malignancies based on the fatty acid indexes and clinical test results improved the diagnostic performance of current clinical level. Then, we used machine learning to define three different FFA metabolic clusters of biliopancreatic malignancies, and survival analysis showed significant differences in prognostic outcomes among the three clusters. This study found metabolic alterations in biliopancreatic malignancies based on bile samples, which ma