AI-Based Approach Diagnoses Colorectal Cancer from Gut Microbiota
Posted on 27 Aug 2025
Colorectal cancer is the second leading cause of cancer death worldwide, and diagnosis is often delayed due to the discomfort and cost of colonoscopies, the current gold standard. Many cases are detected only at advanced stages, when treatment options are limited. With cases also rising among young adults, there is an urgent need for simpler, less invasive diagnostic methods. Researchers have now developed an approach that uses stool samples to detect the disease with high accuracy.
A research team from the University of Geneva (Geneva, Switzerland) has created a machine learning model that analyzes gut microbiota to identify colorectal cancer. Instead of focusing on species or individual strains, the method examines bacterial subspecies, which better capture functional differences linked to disease. The researchers first built a comprehensive catalog of human gut microbiota subspecies, then combined this with clinical data to design an accurate, non-invasive diagnostic tool.
The study, published in Cell Host & Microbe, demonstrated that this artificial intelligence (AI)-based method detected 90% of colorectal cancer cases, a rate close to the 94% achieved by colonoscopies. It outperformed all existing non-invasive methods while using only stool samples. By identifying subspecies-level differences, the approach highlights how gut bacteria functionally contribute to cancer development, offering deeper biological insights.
With further integration of clinical data, the model could eventually match the accuracy of colonoscopy. It has the potential to become a routine, first-line screening tool, allowing colonoscopies to be reserved for high-risk patients. A clinical trial is already being launched in collaboration with Geneva University Hospitals to better determine the cancer stages and lesions detectable with this approach.
The implications extend beyond colorectal cancer, as the same technique could be applied to a wide range of diseases influenced by the microbiota. By distinguishing subspecies-level differences, scientists can uncover mechanisms of action linking gut bacteria to health and disease. This opens possibilities for developing new non-invasive diagnostic tools from a single microbiota analysis.
“Instead of relying on the analysis of the various species composing the microbiota, which does not capture all meaningful differences, we focused on an intermediate level of the microbiota, the subspecies,” said Mirko Trajkovski, professor in the Department of Cell Physiology and Metabolism at the University of Geneva Faculty of Medicine. “The subspecies resolution is specific and can capture differences in how bacteria function and contribute to diseases, including cancer, while remaining general enough to detect these changes across populations.”
