Serum Metabolomic Profiling for Colorectal Cancer using Machine Learning

Abstract

Background: Colorectal cancer is one of the deadliest diseases with a high prevalence worldwide and is characterized by the appearance of adenomatous polyps in the colon mucosa which are at high risk of developing into colorectal cancer. This study aims to use serum metabolites as promising non-invasive biomarkers for colorectal cancer detection and prognostication. Differences in serum metabolites in patients with adenomatous polyps, colorectal cancer, and healthy controls are considered to be able to support the prognosis of colorectal cancer. Methods: Metabolite dataset is taken from the Metabolomic Workbench. Analysis and validation are carried out in silico using machine learning methods. Results: From a total of 234 samples, 113 metabolites were found and 5 metabolites; histidine, lysine, glyceraldehyde, linolenic acid, and aspartic acid were identified as the most significant in differentiating the sample groups. CTD analysis showed that aspartic acid and histidine are associated with the biological pathways of colorectal cancer progression and significant metabolites are associated with cancer-related phenotypes. Conclusion: The serum metabolites differ in colorectal cancer and healthy control. The significant metabolites can be used as a consideration in selecting colorectal cancer biomarkers, but improvisation is needed to obtain more accurate biomarkers.