Scientists from The Ohio State University have developed a machine learning tool to identify metabolic differences between individuals with colorectal cancer and healthy individuals. The tool, which analyzes blood samples, could serve as a noninvasive method for diagnosing and monitoring the progression of the disease. The study was published in the journal iMetaOmics.
The research team analyzed biological samples from over 1,000 people, including 626 colorectal cancer patients and 402 healthy controls. The tool, named PANDA (PLS-ANN-DA), combines two machine learning techniques to distinguish molecular profiles associated with cancer. It also revealed metabolic shifts linked to disease severity and genetic mutations that increase cancer risk.
Dr. Jiangjiang Zhu, co-senior author of the study, emphasized the tool’s potential for monitoring treatment effectiveness. “If metabolites can indicate a treatment’s response faster than traditional methods, they could help doctors adjust therapies more quickly,” Zhu explained. However, the tool is not intended to replace colonoscopy, the current gold standard for cancer screening.
The study leveraged a unique dataset from large research projects, enabling a comprehensive analysis of metabolic changes across different cancer stages. Notably, the team identified heightened activity in purine-related pathways in cancer patients, which may offer clues about underlying disease mechanisms.
While the results are promising, further validation with additional samples is needed before the tool can be applied clinically. The researchers plan to refine the PANDA pipeline by analyzing more metabolic signals to improve its accuracy.
This breakthrough highlights the potential of machine learning in advancing cancer diagnostics. By providing a faster, noninvasive way to detect and monitor colorectal cancer, the PANDA tool could eventually enhance patient care and treatment outcomes. Future research will focus on optimizing the tool for broader clinical use.
Add comment