The trillions of microbes that inhabit the human gut have long been suspected of playing a role in colorectal cancer, but the rectum itself has remained a comparatively understudied corner of that relationship. A new study published in Applied Microbiology and Biotechnology now offers one of the clearest pictures yet of how the gut microbiota changes in patients with rectal cancer, both at the time of initial diagnosis and after treatment, and demonstrates that machine learning algorithms can read those microbial signatures with remarkable accuracy to distinguish cancer patients from healthy individuals. The research, led by a team at Wenzhou Medical University and Zhejiang Tumor Hospital in China, analyzed fecal samples from 77 rectal cancer patients and 83 healthy controls using 16S rRNA gene sequencing, a technique that catalogs bacterial species by reading a stretch of DNA shared across all bacteria but variable enough to tell them apart. The findings suggest that gut microbial profiling could eventually complement blood-based tumor markers as a non-invasive diagnostic and monitoring tool for this disease, which accounts for roughly one third of all colorectal cancers worldwide.
The study was designed around a clinically meaningful distinction that most microbiome studies of colorectal cancer have glossed over: the difference between patients who have just been diagnosed with rectal cancer, known as initially diagnosed rectal cancer or iRC, and those who have already undergone treatment, referred to as post-treatment rectal cancer or tRC. Treatment for rectal cancer typically involves surgery, often combined with radiotherapy and chemotherapy, all of which can profoundly alter the gut environment. By including both groups, the researchers could ask not only whether the microbiome differs between cancer and health, but also whether it recovers after the tumor has been addressed. The answer, it turns out, is nuanced. The microbial community in treated patients showed partial restructuring compared with the newly diagnosed group, but it did not revert to a healthy state, hinting that the disruption caused by cancer and its treatment may leave a durable imprint on the gut ecosystem.
The sequencing data revealed several striking patterns. In newly diagnosed patients, the overall diversity of gut bacteria, a measure known as alpha diversity that reflects how many species are present and how evenly distributed they are, was significantly lower than in healthy controls. Reduced microbial diversity is a recurring theme in colorectal cancer research and is often interpreted as a sign of a destabilized, less resilient ecosystem. Beyond diversity, the composition itself shifted in ways that are biologically suggestive. Bacteria belonging to the genera Bacteroides, Fusobacterium, and Prevotella were enriched in the initially diagnosed patients. Fusobacterium in particular is a well-known character in colorectal cancer biology; the anaerobic bacterium Fusobacterium nucleatum has been shown in numerous studies to promote tumor growth, evade immune responses, and even travel through the bloodstream to colonize metastatic sites. Its enrichment here, along with Bacteroides and Prevotella, reinforces the idea that a specific constellation of potentially pro-inflammatory or pro-tumoral organisms flourishes in the rectal cancer gut.
Just as telling was what was missing. The beneficial bacteria that were depleted in the cancer patients included Faecalibacterium and Subdoligranulum, two genera famed for their role in producing butyrate, a short-chain fatty acid generated when these microbes ferment dietary fiber in the colon. Butyrate is the primary energy source for the cells lining the colon, helps maintain the protective mucus barrier, regulates immune signaling, and generally keeps inflammation in check. A decline in butyrate-producing taxa is therefore thought to remove a layer of protection against carcinogenesis and tumor progression. In this study, the depletion of Faecalibacterium and Subdoligranulum in newly diagnosed rectal cancer patients fits neatly into that framework and offers a mechanistic thread connecting microbial ecology to tumor biology. The fact that these protective organisms remained altered even in treated patients raises the possibility that targeted interventions, such as dietary fiber supplementation, prebiotics, or fecal microbiota transplantation, might one day be used to restore a healthier gut ecosystem in cancer survivors.
To move from observation to practical application, the team built two separate random forest models, a machine learning approach that combines the votes of many decision trees to classify samples based on their microbial profiles. One model was trained to distinguish newly diagnosed patients from healthy controls, and the other to distinguish treated patients from healthy controls. Both models achieved areas under the receiver operating characteristic curves, or AUCs, exceeding 0.9. An AUC of 1.0 represents perfect discrimination, and values above 0.9 are generally considered excellent, meaning the models were able to correctly separate patients from healthy individuals the vast majority of the time based purely on the relative abundances of gut bacterial genera. This level of performance is notable because it was achieved with a relatively modest sample size and with standard 16S rRNA sequencing rather than deeper whole-genome approaches, suggesting that the microbial signals associated with rectal cancer are strong enough to be detected with widely available laboratory techniques.
Perhaps the most clinically compelling finding involved Prevotella. In the initially diagnosed group, the abundance of this genus was positively correlated with serum levels of four established tumor markers: CEA, the classic carcinoembryonic antigen used in colorectal cancer monitoring, along with CA199, CA242, and CA50, a panel of carbohydrate antigens that oncologists routinely track to assess tumor burden and treatment response. A positive correlation between a gut microbe and tumor marker levels implies that the microbe’s abundance rises in step with disease activity, positioning Prevotella as a candidate microbial biomarker of disease status. If validated in larger and longitudinal cohorts, measuring gut microbial markers alongside traditional blood markers could improve diagnostic sensitivity, particularly since blood-based tumor markers like CEA are imperfect, missing a meaningful fraction of early-stage cancers and sometimes elevated in benign conditions.
The study also examined predicted functional profiles of the microbiota, using computational tools that infer what metabolic pathways the bacterial communities are likely performing based on which species are present. These predicted functions differed significantly among the healthy, newly diagnosed, and treated groups, indicating that the compositional shifts translate into changes in the collective metabolic output of the gut ecosystem. While such predictions are indirect and require experimental confirmation, they provide hypotheses about how gut dysbiosis might influence cancer biology, for example through altered production of inflammatory metabolites, changes in bile acid metabolism, or shifts in the fermentation pathways that generate protective short-chain fatty acids.
The researchers emphasize several limitations and cautions inherent in this kind of work. The study is cross-sectional, capturing a snapshot in time rather than following patients longitudinally, which limits the ability to establish whether the microbial changes are a cause or a consequence of the cancer. Sequencing fecal samples captures bacteria shed into the stool rather than those directly attached to the tumor mucosa, and predicted functions are inferred rather than measured. Furthermore, treatments such as antibiotics, radiotherapy, and chemotherapy can independently reshape the microbiome, complicating the interpretation of differences between the iRC and tRC groups. Nonetheless, the consistent high performance of the classification models and the concordance between microbial taxa and clinical tumor markers suggest that the signals are robust and biologically anchored.
The significance of this work lies partly in its focus on rectal cancer specifically, rather than colorectal cancer as an undifferentiated whole. Colon and rectal tumors differ in anatomy, local environment, treatment pathways, and increasingly in epidemiology, with rectal cancer rising in younger adults in many countries. Microbiome studies that lump these sites together may dilute site-specific signals, and the current study demonstrates that rectal cancer leaves its own distinct microbial fingerprint, one that persists in altered form even after treatment. That persistence matters for the growing population of rectal cancer survivors, whose long-term gut health, risk of recurrence, and quality of life may all be intertwined with the state of their microbiota.
Looking ahead, the authors suggest that combining microbial biomarkers with machine learning may provide a non-invasive approach for distinguishing microbial profiles across different clinical groups. Fecal sampling is painless, inexpensive, and easily repeated, making it attractive for screening, monitoring treatment response, and surveillance after therapy. Before such tools enter clinical practice, however, larger prospective studies will be needed to validate the diagnostic models, standardize sequencing and analytical protocols, and test whether microbial markers can predict recurrence or guide therapy decisions in real time. Interventional trials will also be required to determine whether restoring beneficial butyrate-producing bacteria in patients can genuinely influence outcomes. The study was supported by the Natural Science Foundation of Zhejiang Province and the Wenzhou Science and Technology Bureau, and the authors declare no competing interests. For now, the research adds a substantial piece to the growing body of evidence that the gut microbiome is not a passive bystander in rectal cancer but an active participant whose fingerprints can be read, with the help of artificial intelligence, from a simple stool sample.
Subject of Research: Gut microbiota signatures associated with diagnosis and treatment in rectal cancer, analyzed using 16S rRNA sequencing and random forest machine learning models
Subject of Research: Biology
Article Title: Gut microbiota signatures associated with diagnosis and treatment in rectal cancer
Article References: Zheng, X., Zhu, M., Lu, M., Li, X., Liu, Z., Zhou, Y., Mao, C., Du, J., & Xu, C. (2026). Gut microbiota signatures associated with diagnosis and treatment in rectal cancer. Applied Microbiology and Biotechnology. https://doi.org/10.1007/s00253-026-14024-4
Image Credits: AI Generated
DOI: 10.1007/s00253-026-14024-4
Keywords: Rectal cancer, Gut microbiota, 16S rRNA sequencing, Machine learning, Biomarkers, Bacteroides, Fusobacterium, Prevotella, Faecalibacterium, Butyrate, Tumor markers, Dysbiosis
Cite Scienmag News
APA
MLA
Chicago
Morgan Morrow. (September 8, 2026). Gut bacteria linked to rectal cancer diagnosis and treatment outcomes. Scienmag. https://scienmag.com/gut-bacteria-linked-to-rectal-cancer-diagnosis-and-treatment-outcomes/
Morgan Morrow. “Gut bacteria linked to rectal cancer diagnosis and treatment outcomes.” Scienmag, 8 September 2026, https://scienmag.com/gut-bacteria-linked-to-rectal-cancer-diagnosis-and-treatment-outcomes/. Accessed 8 September 2026.
Morgan Morrow. “Gut bacteria linked to rectal cancer diagnosis and treatment outcomes.” Scienmag. September 8, 2026. https://scienmag.com/gut-bacteria-linked-to-rectal-cancer-diagnosis-and-treatment-outcomes/
Copy citation
Download RIS
Tags: 16S rRNA gene sequencing for cancer detection16S rRNA gene sequencing in cancer researchfecal microbial profiling for colorectal cancerfecal microbiota profiling in colorectal cancergut bacteria and rectal cancer treatment outcomesgut bacteria as biomarkers for rectal cancergut microbiota and colorectal cancer risk factorsGut microbiota in rectal cancerGut microbiota in rectal cancer diagnosisimpact of gut bacteria on rectal cancer treatment outcomesmachine learning algorithms in microbiome researchmachine learning for microbial signature analysismachine learning in microbial signature analysismicrobial changes before and after cancer treatmentmicrobial profilingmicrobial signatures distinguishing cancer patients from healthy controlsmicrobiome changes after cancer treatmentmicrobiome signatures distinguishing healthy vs cancer patientsmicrobiome-based cancer diagnosticsmicrobiota-based diagnostic tools for colorectal cancernon-invasive colorectal cancer biomarkersnon-invasive colorectal cancer detection methodsrole of gut microbes in cancer diagnosisrole of gut microbes in rectal cancer progression
