Mouth-to-gut microbial transmission signatures enable robust, non-invasive diagnosis of gastrointestinal cancers.

The human microbiome is spatially compartmentalized, yet oral bacteria can ectopically colonize distal sites such as the gut, potentially influencing disease. By analyzing paired oral and fecal microbiomes from 507 participants across healthy controls and patients with metabolic disorders or gastrointestinal cancers, we established a quantitative mouth-to-feces (MF) index to measure MF microbial transmission. The MF index revealed elevated mouth-to-gut transmission in cancer and a strong association with host metabolic and inflammatory indicators. Using transmitted taxa, we developed a random forest classifier that accurately distinguished gastric/colorectal cancer from healthy controls across seven independent cohorts, even when trained solely on oral microbiome data. When benchmarked against the conventional screening test, the MF-based model achieved markedly higher sensitivity than the fecal occult blood test. These findings uncover disease-specific transmission signatures and highlight MF microbial profiling as a generalizable, non-invasive framework for gastrointestinal cancer diagnosis and risk stratification.