A Blood Test Using Microbiome DNA Could Catch Stomach Cancer Before Symptoms Start

Stomach cancer is the fifth most common cancer worldwide and one of the deadliest, mainly because the vast majority of patients are diagnosed only after the disease has spread. A new liquid biopsy assay published in npj Precision Oncology takes a different angle on the early-detection problem: instead of looking for tumor DNA shed by the cancer itself, it reads circulating microbiome DNA, the fragments of microbial genetic material that float freely in human blood plasma. The results suggest this approach can identify early-stage gastric cancer patients from healthy controls with striking accuracy across independent validation cohorts.

Circulating Microbiome DNA Distinguishes Early Gastric Cancer from Healthy Controls

The central finding is that patterns in cell-free microbial DNA extracted from a standard blood draw can reliably flag gastric cancer at stages when curative surgery is still possible. Across the study's primary cohort and two independent validation cohorts, the assay consistently separated cancer patients from non-cancer individuals with area-under-the-ROC-curve (AUROC) values ranging from 0.937 to 0.972. In plain terms, an AUROC of 1.0 would be a perfect test; values above 0.93 represent strong discriminative performance for a blood-based cancer screen. In the primary study cohort, working at a specificity of 92.1%, the test achieved a sensitivity of 88.2%, meaning it correctly flagged roughly nine in ten early-stage cancer cases while keeping false positives low. The first validation cohort returned similarly strong numbers: 91.5% of healthy individuals and 91.8% of gastric cancer patients were correctly classified.

Why Early Detection of Gastric Cancer Has Been So Hard to Achieve

Those numbers matter because the current diagnostic toolkit for gastric cancer is badly mismatched to the task of population screening. Gastroscopy is the gold standard for confirming gastric cancer, but it is invasive, expensive, and widely avoided by people who are asymptomatic, which is precisely the group that early screening must reach. Outside of Japan and South Korea, no organized national screening programs exist for gastric cancer in the general population, leaving most cases in Europe, North America, and much of Asia to be discovered incidentally at late stages when five-year survival rates drop sharply. A non-invasive blood test that could precede or triage gastroscopy would address all three of those barriers simultaneously: it requires only a venous blood draw, can be performed at scale, and does not depend on patient willingness to undergo an endoscope.

The Microbial Signal Hidden in Plasma

The scientific premise of the assay rests on a body of evidence showing that the gut microbiome differs meaningfully between cancer patients and healthy individuals, and that microbial DNA fragments from those communities enter the bloodstream in detectable quantities. What the research team demonstrated is that the specific composition and abundance patterns of circulating microbiome DNA, analyzed from plasma using next-generation sequencing, carry a cancer-associated signature strong enough to drive a classifier. The assay does not rely on a single microbial species or biomarker; instead, a machine learning model trained on the multi-dimensional microbial profile assigns each sample a probability score. The approach fits within the broader shift toward liquid biopsy as a minimally invasive oncology tool, where the goal is extracting actionable clinical information from blood rather than tissue.

A particularly notable finding involves anatomical coverage. The assay detected tumors located in the gastric cardia and fundus, regions that are notoriously difficult to visualize cleanly with standard gastroscopy because of their position near the gastroesophageal junction. The test's performance was not confined to easier-to-detect tumor locations, which strengthens confidence that the microbial signal is systemic rather than site-specific.

Limitations and the Validation Gap That Still Needs to Close

The study's honest accounting of limitations deserves emphasis. Sample sizes across the three cohorts are in the hundreds rather than thousands, which means the performance estimates, however encouraging, carry uncertainty intervals that only a larger prospective trial can narrow. The cohorts were drawn from Chinese hospital populations, a demographic in which gastric cancer rates and microbiome composition may differ from populations in other regions, raising questions about generalizability. The NCI and other regulatory bodies note that liquid biopsy tests for healthy-population screening require extensive additional evidence before they can reliably substitute for established screening methods. The published data are a proof-of-concept and a strong one, but they are not yet sufficient to guide clinical guidelines or reimbursement decisions. Whether the specific microbial taxa driving the classifier reflect a causal relationship with gastric carcinogenesis or are simply a correlated bystander signature also remains to be determined.

The Road to Clinical Deployment for a Microbiome-Based Gastric Cancer Screen

The immediate next steps are predictable and necessary: a larger multi-center prospective cohort, ideally spanning populations with different baseline microbiome compositions and different gastric cancer incidence rates, is needed to establish the assay's real-world performance outside the controlled hospital setting. Comparisons against existing non-invasive candidates, including serum pepsinogen ratios and Helicobacter pylori serology already in use in parts of Asia, will be needed to determine where a microbiome DNA test fits in a practical screening algorithm rather than simply whether it works in isolation. If the assay's sensitivity and specificity hold in those larger studies, non-invasive microbiome-based screening could plausibly serve as a first-pass filter that reserves gastroscopy for patients who test positive, improving both the uptake and the cost-effectiveness of population screening programs globally. The biology is compelling, the initial numbers are strong, and the unmet clinical need is enormous. The work that remains is the hard, slow business of validation at scale.