In a paper published in Frontiers in Oncology, researchers analysed one of the largest UK Biobank dataset of protein profiles from healthy individuals and colorectal cancer patients and highlighted three proteins - TFF3, LCN2, and CEACAM5 - as important markers linked to cell adhesion and inflammation, processes closely associated with cancer development. The next steps would require further validation of these biomarkers and then they may be developed into new diagnostic tools.
Three different machine learning models and artificial intelligence (AI) are used to recognise patterns in data.
Dr Animesh Acharjee, from the Department of Cancer and Genomic Sciences & Deputy Programme Director, MSc in Health Data Science (Dubai) who led the study said:
"Colorectal cancer is a leading cause of cancer-related deaths worldwide and it is predicted to increase in incidence over coming decades. This increase highlights the need for reliable tools to diagnose and predict the disease, especially since earlier detection allows for more effective treatment.
"This study results offer valuable insight for identifying potential biomarkers in future proteomic studies and it is hoped this knowledge will eventually help improve treatments for patients with colorectal cancer.
"In our study, we used advanced machine learning and artificial intelligence (AI) models combined with protein network analysis to identify key protein biomarkers that could aid in diagnosing colorectal cancer. The biomarkers show promise but further large-scale validation study is needed to look into the relationships and mechanistic properties of these potential new biomarkers."
Colorectal cancer is the fourth most common cancer in the UK, with around 44,100 people are diagnosed each year. This type of cancer occurs when abnormal cells start to divide and grow in an uncontrolled way, affects the large bowel, which is made up of the colon and rectum.
Currently, diagnosis involves a doctor removing tissue from the bowel and sending a sample of cells to the laboratory for various tests that can identify cancer and indicate which treatments may work best. Any advances that can help pick up colorectal cancer sooner and in a way that is more straightforward for patients would be welcomed.
Radhakrishnan SK, Nath D, Russ D, Merodio LB, Lad P, Daisi FK, Acharjee A.
Machine learning-based identification of proteomic markers in colorectal cancer using UK Biobank data.
Front Oncol. 2025 Jan 7;14:1505675. doi: 10.3389/fonc.2024.1505675