Invention of a Test That Diagnoses Colon Cancer... Using Artificial Intelligence
Researchers at the University of Geneva in Switzerland have developed a non-invasive diagnostic method that could revolutionize the early detection of colorectal cancer, moving away from invasive colonoscopies. The approach relies on analyzing simple stool samples rather than the traditional, costly, and uncomfortable endoscopic procedure, which is often cited as a reason why many people delay getting screened.
The researchers explained that they used machine learning techniques to create the first detailed catalog of human gut bacteria at a level of precision that allows for an understanding of how different sub-microbial groups function within the body. They then leveraged this information to detect colorectal cancer based on the bacteria present in stool samples, providing a low-cost and non-invasive alternative. The study’s findings were published in the journal *Cell Host & Microbe*.
Lead researcher Mirko Trajkovski, a professor in the Department of Cell Physiology and Metabolism at the University of Geneva Faculty of Medicine, noted that the team focused on an intermediate level of microbial classification known as “subspecies,” rather than analyzing full bacterial species, which fails to capture subtle differences, or strains, which vary significantly between individuals. He explained that this intermediate level enables the monitoring of functional differences between bacteria and their relationship to diseases, including cancer, while maintaining a degree of generality that allows these differences to be observed across different populations or countries.
For his part, Mattia Trejkovic, a PhD student in the research team and the study’s lead author, clarified that the biggest challenge was developing an innovative method to analyze massive amounts of biological data. He emphasized that the team succeeded in building the first comprehensive catalog of human gut microbial subspecies, along with an accurate and effective method for using it in both research and clinical applications.
• The model successfully detected 90 percent of cancer cases, a rate very close to the 94 percent detection rate achieved by traditional colonoscopy.
• The model’s accuracy surpassed all currently available non-invasive detection methods, paving the way for its use as a primary routine screening test, after which positive patients would be referred for colonoscopy solely for confirmation.
The researchers confirmed that additional clinical data could improve the model’s accuracy in the future, bringing its performance closer to that of traditional colonoscopy. Preparations are currently underway for a clinical trial in collaboration with University of Geneva hospitals to determine the stages and cancerous lesions that this method can detect with greater precision.
Trajkovski concluded that the same approach could be used in the future to develop non-invasive diagnostic tools for a wide range of other diseases, relying on a single analysis of gut microbes.
