University of Geneva Develops Stool Test Detecting 90% of Colorectal Cancers Using Machine Learning

A Breakthrough in Non-Invasive Screening

Researchers at the University of Geneva (UNIGE) have developed a stool-based diagnostic tool that identifies 90% of colorectal cancer cases using machine learning to analyze gut microbiota at the subspecies level. Published in Cell Host & Microbe, the method leverages a detailed catalog of human gut microbiota subspecies to detect cancer-related patterns in stool samples. This approach outperforms existing non-invasive methods like FIT (Fecal Immunochemical Test) and Cologuard, which typically achieve 70-80% sensitivity. By examining microbial subspecies rather than broader species or strains, the test captures functional differences in bacteria that may signal cancer progression. This innovation eliminates the need for colonoscopies, which are invasive, costly, and often uncomfortable, addressing a key challenge in early detection.

Microbial Subspecies and Cancer Risk

The study’s lead investigator, Mirko Trajkovski, emphasized that subspecies-level analysis provides higher microbial activity resolution, enabling the identification of cancer-promoting bacterial strains. Certain subspecies of Fusobacterium and Enterococcus have been linked to tumor development, while others may inhibit carcinogenic processes. By mapping these interactions, the test generates a predictive model linking microbiota composition to cancer risk. This innovation could enhance screening protocols, particularly for younger demographics where colorectal cancer incidence is increasing. However, the test’s ability to detect precancerous lesions like advanced adenomas remains unproven, a limitation shared with other stool-based methods.

Challenges to Widespread Adoption

“subspecies-level analysis provides higher microbial activity resolution, enabling the identification of cancer-promoting bacterial strains.”

Despite its potential, the UNIGE test faces several challenges before widespread adoption. Its 90% sensitivity is slightly lower than colonoscopy’s 94% detection rate, the current gold standard for diagnosing colorectal cancer. This accuracy means the test may miss approximately 10% of cases, requiring confirmatory colonoscopies for positive results. Additionally, the test’s capacity to identify precancerous polyps or early-stage lesions is unvalidated. Current stool tests, including the UNIGE method, struggle to detect small, early-stage tumors, which are critical for preventing cancer progression. This limitation underscores the need for complementary diagnostic tools.

Regional Variations and Technical Barriers

Another concern is the test’s reliance on a database of gut microbiota subspecies, which may not account for regional variations in microbial composition. Gut flora in populations from different geographic regions or with distinct dietary habits may differ significantly, potentially affecting the test’s generalizability. Furthermore, the method requires advanced AI and sequencing technologies not yet standardized for routine clinical use. While the UNIGE team claims the test is cost-effective and accessible, its practical implementation depends on overcoming these technical and logistical hurdles.

Clinical Validation and Future Applications

To address these limitations, UNIGE researchers are conducting clinical trials at Geneva University Hospitals (HUG) to refine the test’s ability to detect cancer stages and lesion types. The trial aims to evaluate the method’s performance in identifying advanced adenomas, polyps, and early-stage tumors, which are essential for preventive care. However, the trial’s results are not yet publicly available, and experts caution that large-scale validation across diverse populations is necessary to confirm the test’s reliability.

Expanding Beyond Colorectal Cancer

The study also explores the potential of microbiota analysis for diagnosing other diseases beyond colorectal cancer. Similar approaches could be applied to detect pancreatic, lung, or breast cancers by identifying microbial signatures associated with these conditions. However, such applications require extensive research to establish microbial markers for each disease. The UNIGE team acknowledges that while the current method represents a significant advancement, its broader implications depend on further scientific validation and regulatory approval.

Practical Implementation Challenges

The stool test’s practical implementation faces several challenges, including sample collection, data interpretation, and patient compliance. As a non-invasive method, it requires patients to provide stool samples, which may be less intimidating than colonoscopies but still pose logistical hurdles. Improper sample storage or labeling could compromise test results, a problem observed in other stool-based tests where over 10% of samples are deemed unsatisfactory. Additionally, the test’s accuracy may vary depending on factors like diet, antibiotic use, or gut microbiota composition, which could affect its consistency across populations.

Integration into Healthcare Systems

“while the current method represents a significant advancement, its broader implications depend on further scientific validation and regulatory approval.”

Integration of this technology into existing healthcare systems also presents barriers. While the test offers a low-cost alternative to colonoscopies, its adoption would require training healthcare providers to interpret results and manage follow-up procedures. Insurance coverage and regulatory approval are critical for widespread use. The UNIGE team is working with regulatory bodies to ensure the test meets safety and efficacy standards, but the timeline for approval remains uncertain. Until these challenges are resolved, the test will likely serve as a complementary tool rather than a replacement for colonoscopies.

Foundational Research and Unpublished Details

A key component of the UNIGE study was the work of Matija Trickovic, who developed tools to catalog bacterial subgroups and apply them to medical testing. Trickovic’s research provided the foundational data for mapping gut microbiota subspecies, enabling the team to create a comprehensive catalog used in the machine learning model. His methods allowed for the identification of microbial subgroups influencing health outcomes, which was critical for the test’s development. However, the specific details of Trickovic’s study remain unpublished, and further research is needed to validate his findings in broader contexts.

The Future of Microbiome-Driven Diagnostics

The UNIGE study highlights the growing role of the gut microbiome in cancer detection and prevention. By revealing how microbial subspecies influence disease progression, the research opens new avenues for personalized medicine. Future studies could explore how dietary interventions or probiotics might modulate gut bacteria to reduce cancer risk. Additionally, the use of machine learning in microbiome analysis sets a precedent for applying AI to other complex biological systems, such as the gut-brain axis or immune responses.

While the stool test represents a major advancement, its success depends on overcoming current limitations and demonstrating consistent performance in diverse populations. If validated, it could significantly improve colorectal cancer screening rates, particularly among underserved groups facing barriers to colonoscopies. However, until clinical trials confirm its efficacy, the test remains a promising but unproven innovation in the fight against cancer.