TOPLINE

Oral bacteria colonizing the gut — quantified using a new “mouth-to-feces” (MF) transmission index — were significantly more abundant in patients with gastric and colorectal cancer, enabling diagnostic classifiers that outperformed standard fecal occult blood testing (FOBT). The signature was validated across some independent cohorts and remained effective using oral samples alone, pointing to a promising noninvasive alternative to conventional stool-based cancer screening.

METHODOLOGY

• Although oral bacteria are known to colonize the gut in some disease states and have been identified in colorectal tumors, whether oral-to-gut microbial transmission can be quantified in a clinically meaningful way and leveraged for noninvasive gastrointestinal (GI) cancer detection remains unclear.
• Researchers analyzed paired oral and fecal microbiome samples from 507 participants in South Korea, including healthy control individuals (n = 129), individuals with metabolic disorders (metabolic syndrome, hypertension, hyperlipidemia, or type 2 diabetes; n = 215), patients with gastric cancer (n = 77), and patients with colorectal cancer (n = 86), to develop a quantitative metric measuring oral-to-gut bacterial transmission and assess its potential as a noninvasive signature for gastric and colorectal cancer.
• They sequenced 1010 oral and fecal samples in total using 16S ribosomal RNA gene sequencing, including 502 participants who provided matched oral-fecal pairs, to identify shared bacterial DNA signatures between the mouth and gut and to develop an MF transmission index that quantified the proportion of the gut microbiome comprising bacteria detected in the mouth.
• Machine learning models were built based on MF transmission patterns and validated across seven independent cohorts to determine whether oral and fecal microbial signatures, including models derived from oral samples alone, could distinguish gastric and colorectal cancers.
• Model performance was subsequently compared with FOBT results in an external cohort of 66 participants, including 28 patients with colorectal cancer and 38 healthy control individuals.

TAKEAWAY

• The mean MF index was significantly higher in patients with GI cancers (gastric cancer, 3.63; colorectal cancer, 3.09) than in healthy control individuals (1.19), corresponding to approximately 2.6- to 3.0-fold higher oral-to-gut microbial transmission. By contrast, no significant difference was observed between healthy control individuals and individuals with metabolic disorders.
• A machine learning classifier based on MF-transmitted bacteria detected colorectal cancer from both oral and fecal samples (area under the curve [AUC], 0.82 for each), with somewhat lower performance for gastric cancer (AUCs, 0.66 for oral samples and 0.73 for fecal samples). Performance was maintained when validated in independent international cohorts (AUC range, 0.71-0.85 for colorectal cancer and 0.61-0.65 for gastric cancer), supporting generalizability; accuracy was consistently lower for early-stage disease than for late-stage disease.
• In a direct comparison between MF-based models using oral and fecal microbiome samples and FOBT, the models distinguished colorectal cancer with a sensitivity of 92.9% and specificities of 92.1% and 94.7% for oral and fecal samples, respectively (AUCs, 0.969 and 0.984, respectively), whereas FOBT had a sensitivity of 61.0% and a specificity of 100% (AUC, 0.80).
• Oral sample accuracy held up in external validation only when bacteria were identified down to their exact genetic sequence, rather than grouped by genus; collapsing to genus-level groupings degraded oral classification performance across external cohorts (eg, colorectal cancer accuracy fell from AUC 0.82 to 0.33 in one validation cohort). Stool-based testing did not share this limitation.

IN PRACTICE

“These insights provide a deeper understanding of the role of oral microbiota in cancer progression and suggest the potential utility of MF microbial profiles in cancer staging, early detection, and microbiome-informed therapeutic strategies,” the authors of the study wrote.

In a related press release, it was noted that “if the findings are validated with further studies, the researchers say the next step would be developing a noninvasive test. An oral-rinse test, in particular, would be easy to collect and generally less burdensome than a stool test, the authors note, and could help identify people who should be prioritized for established diagnostic procedures such as upper endoscopy or colonoscopy.”

SOURCE

The study was co-first authored by Lae-Guen Jang and Ji-Won Huh, with lead contact Jihyun F. Kim of Yonsei University in Seoul, Republic of Korea. It was published online on August 20, 2026, in Cell Host & Microbe.

LIMITATIONS

The study did not determine whether increased oral-to-gut microbial transmission drove cancer progression or reflected cancer-related microbial changes. The analysis did not fully distinguish bacterial strains, and genetically similar bacteria detected in both the mouth and gut may have differed in their biologic functions. The findings required broader validation across diverse populations because differences in sequencing methods, sample handling, diet, medication use, and immune status may have affected model performance.

DISCLOSURES

The study received financial support from the National Research Foundation of Korea and the Yonsei Signature Research Cluster Program; two authors also disclosed partial support from Basgen Bio and the Research Initiative Program at Chungnam National University, respectively. Some authors disclosed being listed as inventors on a patent application on MF profile-based diagnostics that is to be filed.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.