Large-Scale Gut Microbiome Analysis Identifies a Robust Colorectal Cancer Signature

The gut microbiome has long been suspected of playing a role in colorectal cancer, but prior studies were often small, used different sequencing methods, and produced results that were difficult to compare. A new study published in Cell Host & Microbe addresses this directly, reanalyzing data from 27 independent studies comprising 6,779 publicly available gut microbiome sequencing profiles to identify a robust microbial signature associated with the disease.

A central methodological contribution was a machine-learning algorithm trained to distinguish cancer from non-cancer microbiomes. "It outputs a score of how 'cancer-like' a microbiome is," said corresponding author Georg Zeller from EMBL Heidelberg. "We can apply this to any existing human gut microbiome dataset, including from dietary intervention studies." The resulting signature was not limited to one cohort, geography, sequencing method, or age of diagnosis, and appeared consistently across both early-onset and late-onset cases. 

The study also analyzed 906 intestinal tissue samples to compare stool-based signals with microbes found directly in tumor tissue. The two largely mirrored each other, and cancer-associated microbes were detectable even in early-stage tumors. Detection accuracy in stool was somewhat lower for early-stage cancers and tumors located further from the rectum, possibly because these tumors are smaller or more distant from the point of sample collection.

Pre-cancerous adenomas proved harder to detect. Adenoma-associated microbial changes were weaker and showed limited overlap with the cancer signature. "It suggests that more sensitive approaches, larger datasets, or combinations with other measurements may be needed before microbiome-based tools could contribute to the reliable detection of early pre-cancerous lesions," said co-author Michael Zimmermann.

Diet also emerged as a relevant variable. A stronger cancer-associated microbiome score was linked to lower dietary fiber intake, and dietary intervention studies showed that increasing fiber consumption was associated with a reduction in that score.

The study further examined Fusobacterium, a bacterial group repeatedly linked to colorectal cancer, and found meaningful differences between subspecies. Fusobacterium nucleatum subsp. animalis showed consistent enrichment across continents, while other subspecies showed more geographically variable patterns, with several found almost exclusively in cancer patients from Asia.

The researchers emphasize that these findings do not constitute a diagnostic test—microbiome-based classifiers did not yet match the performance of fecal immunochemical tests in head-to-head comparisons. The study's value lies in defining a reproducible disease-associated signature from publicly available data, providing a foundation for future work in microbiome-based risk assessment and early detection.