Biomarker Metabolite Discovery for Pancreatic Cancer using Machine Learning

Abstract

Pancreatic cancer is one of the deadliest cancers in the world. This cancer is caused by multiple factors and mostly detected at late stadium. Biomarker is a marker that can identify some diseases very specific. For pancreatic cancer, biomarker has been recognized using blood sample known as liquid biopsy, breath, pancreatic secret, and tumor marker CA19-9. Those biomarkers are invasive, so we want to identify the disease using a very convenient method. Metabolite is product from cell metabolism. Metabolites can become a biomarker especially from difficult diseases. In this paper, we want to find biomarker from metabolite using machine learning and enrichment. Metabolites data was obtained from Metabolomic workbench, while the detection and identification is done using in silico. From 106 samples, control and cancer, we found 61 metabolites and analyze them. We got 8 metabolites that play important role in pancreatic cancer and found out 2 of them are the most impactful. From that we found that ethanol is one of the best candidate of biomarker that we provide for pancreatic detection cancer. However, the simulation need to be improved to find another biomarker that provide a better marker for prognosis.