Gut Microbiome Science: How Digestive Health Research Advances

Fifteen years ago, studying the human gut microbiome required invasive sampling, weeks of laboratory processing and analytical methods that felt closer to archaeology than to medicine. Gut microbiome research has since moved from academic curiosity toward clinical relevance, and the tools now available bear little resemblance to what came before. What is most striking is not any single discovery but how they are converging into something more coherent: a functional, actionable understanding of the microbial ecosystem that shapes digestive health.

This article surveys where gut microbiome research currently stands, the technological innovations reshaping how microbial communities are monitored and analyzed, and the emerging model of continuous, non-invasive monitoring. It also tries to be explicit about methodological rigor and about the limitations the field still faces.

The Multi-Omics Shift

Early microbiome science had one primary lens: taxonomy. Sequencing 16S ribosomal RNA genes answered a relatively simple question about which bacteria were present. That was revolutionary at the time and now feels narrow.

The scale of what followed is easy to underestimate. The MetaHIT consortium's first gut gene catalogue, built from 124 European individuals, described 3.3 million non-redundant microbial genes, roughly 150 times the number in the human genome (Qin et al., Nature, 2010). Four years later an integrated catalogue built from 1,267 samples across three continents expanded that to 9.9 million genes (Li et al., Nature Biotechnology, 2014). The Human Microbiome Project, working from 242 healthy adults sampled at up to 18 body sites, established the reference picture of healthy variation those catalogues sit within (Nature, 2012).

Modern gut microbiome research integrates several complementary analytical domains. Metagenomics describes not just which organisms are present but their complete genetic potential. Metabolomics quantifies the actual end-products of microbial metabolism, including short-chain fatty acids, secondary metabolites and volatile organic compounds. Proteomics identifies which proteins are actively expressed. Volatile organic compound analysis captures gaseous emissions that reflect real-time microbial activity.

Method	What it measures	Resolution	Functional information	Practical constraints
16S rRNA amplicon sequencing	Taxonomic membership from a marker gene	Genus, sometimes species	Inferred only	Cheap and fast; primer and region choice materially affects results
Shotgun metagenomics	Full genetic content of the community	Species and strain	Genetic potential, not expression	Higher cost and compute; needs deep reference catalogues
Metatranscriptomics	Which genes are transcribed	Species and pathway	Expression, close to real time	RNA is unstable; demanding sample handling
Metabolomics (LC-MS)	Small molecules present in the sample	Compound level	Actual metabolic output	Broad coverage but laboratory-bound; annotation is hard
Volatile organic compound analysis	The volatile fraction of the metabolome	Compound or pattern level	Actual metabolic output, near real time	Volatile subset only; highly sensitive to collection and storage

The value of combining these layers is that each corrects for the blind spots of the others. Genetic potential is not the same as expressed function, and expressed function is not the same as measured metabolic output. The integrative Human Microbiome Project made this concrete by following 132 individuals with and without inflammatory bowel disease for a year with coordinated metagenomic, metatranscriptomic, metabolomic and serological sampling, and showing that multi-omic signatures separated disease activity states more cleanly than composition alone (Lloyd-Price et al., Nature, 2019; Proctor et al., Nature, 2019). Work on data integration in the gut microbiome reaches the same conclusion computationally, reporting that combining taxonomic, genetic and metabolomic layers outperforms any single layer (Li et al., Microbial Cell Factories, 2022).

Precision Nutrition: From Population Averages to Individual Biology

For decades nutritional science operated on population averages. These recommendations miss something fundamental, which is that your microbiome is not identical to anyone else's and your nutritional responses reflect that.

The clearest evidence remains the Israeli personalized nutrition work. Zeevi and colleagues continuously monitored blood glucose in 800 participants across 46,898 meals over a week each, alongside microbiome profiling, and found that postprandial glycemic responses to identical foods varied enormously between people. A machine learning model integrating microbiome features with dietary and clinical data predicted individual responses substantially better than carbohydrate counting, was validated in a separate 100-person cohort, and then tested in a randomized crossover dietary intervention in 26 people, where personalized diets produced lower postprandial glucose than expert-designed ones (Cell, 2015).

That result has since been replicated at scale. The PREDICT 1 study measured postprandial glucose, triglyceride and insulin responses in 1,002 twins and unrelated healthy adults in the United Kingdom, with a further 100-person United States validation cohort, and found that genetics explained only a modest share of variability while meal composition, meal context and the microbiome accounted for much more (Berry et al., Nature Medicine, 2020). A companion analysis of 1,098 deeply phenotyped PREDICT participants linked specific microbial species to habitual diet and to cardiometabolic markers (Asnicar et al., Nature Medicine, 2021).

What remains unsettled is the step from that biology to a reliable commercial recommendation. Two people eating identical diets can experience meaningfully different metabolic outcomes. Whether any given service can identify the right personalized diet from a stool sample is a separate and much less well-evidenced question.

Sequencing: Deeper Resolution, Richer Data

The technical foundations continue to advance rapidly. Sequencing costs have fallen by orders of magnitude over two decades, and long-read technologies have extended read lengths dramatically. Longer reads matter because they allow near-complete microbial genomes to be assembled from complex environmental samples.

Historically, analyzing a fecal sample produced thousands of fragmented sequences from unknown sources. Metagenomic assembly algorithms can now reconstruct high-quality genomes directly from samples, and large collaborative efforts have expanded the catalogue of metagenome-assembled genomes from the human gut substantially. That expanded reference base translates directly into better identification and characterization of the organisms inhabiting the digestive tract.

Volatile Organic Compounds as a Non-Invasive Monitoring Paradigm

Here is what makes VOC analysis compelling: your microbiome is continuously announcing itself. As bacteria ferment dietary components they release volatile organic compounds, gaseous metabolic byproducts that reflect both identity and activity.

Unlike sequencing, which provides a compositional snapshot, volatile organic compound analysis captures functional activity in near real time. The same bacterial species can produce different VOC profiles depending on what it is fermenting, its metabolic state, and its interactions with neighbors.

The evidence base is now large enough to summarize quantitatively. A 2024 systematic review and meta-analysis in the Journal of Crohn's and Colitis reviewed 16 VOC studies in inflammatory bowel disease and meta-analyzed 10 of them, covering 696 cases against 605 controls. Pooled sensitivity was 87 percent (95 percent CI 0.79 to 0.92), pooled specificity 83 percent (95 percent CI 0.73 to 0.90), and the area under the curve 0.92 (Krishnamoorthy et al., 2024). That is a meaningful result and it is also pooled across heterogeneous methods and modest individual cohorts, which is exactly the limitation the field has to work through. For a fuller treatment of this science, see our overview of gut microbiome science and VOC analysis.

From Laboratory Instruments to Accessible Devices

For decades, analyzing volatile organic compounds required gas chromatography-mass spectrometry instruments priced well beyond individual reach and operable only by trained staff. Miniaturization is changing that. Metal oxide sensor arrays combined with machine learning can now detect and classify complex VOC patterns, operating at low cost with minimal consumables.

That enables a shift from intermittent clinical testing toward continuous, passive at-home monitoring. It is worth stating plainly that this remains an emerging capability. What has not yet happened is the large-scale, multi-site clinical validation that would move these from research findings to approved diagnostics.

An adjacent modality shows what completing that path looks like. Zheng and colleagues assembled 5,979 fecal metagenomes across geographies and ethnicities, selected ten bacterial species for an ulcerative colitis model and nine for a Crohn's disease model, reached areas under the curve above 0.90 in the discovery cohort, held performance across trans-ethnic validation cohorts from eight populations, and converted the signature into a droplet digital PCR assay suitable for a clinical laboratory (Nature Medicine, 2024). No fecal VOC test has yet completed an equivalent sequence.

Machine Learning and High-Dimensional Data

Modern gut microbiome research generates data that exceeds human pattern recognition. A single sequencing run produces tens of thousands of reads across hundreds of taxa, and the reference gene catalogue those reads map to now contains 9.9 million genes. Adding metabolomic data introduces thousands more variables. This is not hype about machine learning; it is a practical necessity, because the dimensionality of the data exceeds what twentieth-century statistical approaches handle well.

The more promising approaches use architectures designed for sequential and relational data. Recurrent networks with attention mechanisms can model temporal dynamics of microbiome composition. Graph neural networks capture ecological interactions between taxa. Transformer models developed for language are being adapted to interpret microbiome data (Przymus et al., Frontiers in Microbiology, 2025). Comparative work suggests deep learning can outperform traditional statistical approaches on some microbiome prediction tasks, though results vary considerably by dataset and the size of the improvement should not be overstated.

The External Validation Problem

The single most important caveat in this area is that models often do not survive contact with a new population. Wirbel and colleagues assembled 969 fecal metagenomes across eight colorectal cancer cohorts and showed that classification accuracy measured within a single study systematically overstates what happens when the same model is applied to a different cohort, and that training across studies recovers much of the loss (Nature Medicine, 2019).

The reason is partly biological. Duvallet and colleagues re-analyzed 28 published case-control gut microbiome studies spanning ten diseases with a standardized pipeline and found that roughly half of the genera associated with any individual disease also responded to at least one other disease, meaning many published associations reflect a shared, non-specific response to illness rather than a disease-specific signature (Nature Communications, 2017). Topcuoglu and colleagues make the methodological version of the same argument, documenting how frequently microbiome classification studies conflate model selection with model evaluation (mBio, 2020).

Explainability

High accuracy means little without explainability. A model that predicts a disease state accurately but opaquely is a research curiosity, not a clinical tool. SHAP values, attention mechanisms and integrated gradients allow identification of which specific features drove a prediction. This transforms a black-box output into something clinically actionable: not simply that a profile appears dysbiotic, but that the pattern is characterized primarily by reduced butyrate-producing capacity.

From Reactive to Predictive

Traditional medical testing is reactive. Symptoms appear, testing confirms a diagnosis, treatment follows. Microbiome science raises the possibility of something different: identifying health trajectories before symptoms emerge.

There is at least one longitudinal result that supports the premise. Wilmanski and colleagues analyzed gut microbiome, phenotypic and clinical data from more than 9,000 individuals across three independent cohorts and found that increasing microbiome uniqueness in later life was associated with healthy ageing and, in the oldest participants, with survival over a four-year follow-up (Nature Metabolism, 2021). That is an association in observational data rather than a validated predictive test, and the authors frame it that way.

Research has similarly associated certain microbial depletion patterns, particularly loss of butyrate-producing taxa alongside elevated proteolytic activity, with later development of digestive disorders. This is a genuinely promising direction, but prediction windows in this area are not yet clinically validated, and specific hazard estimates should be treated with caution until confirmed in large prospective cohorts.

Hypothetical scenario. Consider a hypothetical case that illustrates why trajectory matters more than a snapshot. Someone completes a course of broad-spectrum antibiotics. A single stool test six weeks later returns a composition within the normal reference range. Under the mechanisms described in the probiotic reconstitution literature, that same person could still be in a state where mucosal recolonization is incomplete and where metabolic output has not returned to their own pre-antibiotic baseline, because a population reference range is wide enough to hide a personal deviation. This is an illustrative construction, not a real case and not a SNIFR result.

Continuous Monitoring as a New Paradigm

Wearables measure heart rate, skin temperature and activity continuously. The argument for continuous microbiome monitoring is structurally similar. Rather than a snapshot at a clinic visit, an individual could observe the evolution of their microbial ecosystem over time, identifying subtle shifts that a single measurement would miss.

This requires rethinking what gets measured. Comprehensive taxonomic profiling remains expensive and slow. Continuous monitoring might instead focus on a limited set of high-value functional biomarkers: key short-chain fatty acid producers, inflammatory signatures and diversity metrics, with models translating those simplified measurements into a broader assessment.

Clinical Trials and the Evidence Base

The number of registered clinical trials involving microbiome interventions has grown substantially, and more importantly those trials are increasingly focused on clinically meaningful outcomes rather than surrogate markers.

Recurrent Clostridioides difficile infection is the clearest case of that maturation. In a randomized, double-blind trial of 46 patients with three or more recurrences, Kelly and colleagues showed donor fecal microbiota transplantation delivered by colonoscopy outperformed autologous transplantation for resolution of diarrhoea over eight weeks (Annals of Internal Medicine, 2016). The field then moved to a standardized oral product: in the ECOSPOR III phase 3 trial reported by Feuerstadt and colleagues, recurrence at week 8 was 12 percent with the oral spore-based microbiome therapeutic SER-109 compared with 40 percent on placebo, corresponding to sustained clinical response in 88 percent versus 60 percent (New England Journal of Medicine, 2022). That is a defined, randomized, clinically meaningful endpoint, and it is the exception rather than the rule in microbiome therapeutics.

Personalized Probiotic Approaches

One promising application is rational, individualized probiotic design. Rather than off-the-shelf consortia selected by tradition or marketing, the goal is identifying which strains might plausibly benefit a particular individual's microbiome.

The case for personalization here is unusually well documented, and it is largely a negative result about generic products. Zmora and colleagues gave healthy volunteers an eleven-strain probiotic and used upper and lower endoscopy to sample the mucosa directly, finding that gut mucosal colonization was person-specific: some individuals permitted colonization and others resisted it entirely, and stool shedding did not predict which (Cell, 2018). In a companion study, Suez and colleagues found that after antibiotics the same probiotic actually delayed return of the native microbiome to baseline, while autologous fecal transplant restored it within days (Cell, 2018).

Rational strain selection therefore depends on understanding microbial interactions: adding a strain that will be outcompeted by dominant organisms is futile, while adding one that fills a depleted functional niche makes biological sense. Early clinical work in this area is promising but the effect sizes reported vary widely.

Regulatory Landscape

Regulators have historically classified microbiome-based therapeutics inconsistently, sometimes as drugs and sometimes as supplements, which created a complicated environment for developers. That has begun to change with the approval of live biotherapeutic products for recurrent C. difficile infection in the United States, and clearer regulatory pathways generally accelerate clinical development because companies can plan with more certainty.

Digital Biomarkers, Privacy and Ethics

An underappreciated shift is the move from analog clinical assessment to quantified digital biomarkers. Unlike genetic markers, which are static, microbiome composition reflects real-time status and lifestyle, changing over days to weeks. That sensitivity is what makes it useful as a feedback loop.

Microbiome data is also deeply personal. It can reveal not only disease susceptibilities but dietary patterns, cultural practices and details about physiology. As continuous monitoring becomes more common, robust data governance matters: who owns the data, how long it is retained, and under what circumstances it may be shared with third parties such as insurers or employers. Several jurisdictions are extending existing frameworks for sensitive health and genetic data to cover microbiome data, and that protection is essential for building public trust.

Democratization and Its Risks

Gut microbiome research once belonged almost entirely to academic institutions and specialized clinical centers. Technology and business model innovation have lowered the barriers substantially, with direct-to-consumer services and cloud-based analysis platforms making microbiome analysis far more accessible than a decade ago.

This has a dual nature. More people can understand and monitor their own gut health, which is genuinely valuable. It also creates space for misinformation and for products whose claims outrun their evidence. When a company publishes its analytical methods and validation results in peer-reviewed venues, that demonstrates rigor. When claims appear without that foundation, skepticism is warranted. A 2018 review in the BMJ remains a useful reference point for separating what the evidence supports from what is being sold (Valdes et al., 2018).

Standardization

As the field matures, standardization becomes increasingly important, and the scale of the problem has been measured. The Microbiome Quality Control project consortium distributed blinded, identical fecal specimens to 15 laboratories and found that the choice of DNA extraction protocol and bioinformatic pipeline introduced variation comparable in magnitude to real biological differences between people (Sinha et al., Nature Biotechnology, 2017). That is the single most important reason cross-study comparison in this field is difficult.

Standardization initiatives are working toward reference standards, validated protocols and quality assurance frameworks. As those are adopted, cross-laboratory comparability improves and meta-analyses become more meaningful.

Convergence and the Research Agenda Ahead

The future of personalized health monitoring will increasingly integrate microbiome data with other information streams. Genetic testing reveals predispositions; microbiome analysis reveals current functional phenotype; metabolomic profiling reveals metabolic state. Models trained on integrated datasets should have greater predictive power than any single modality.

VOC analysis fits this convergence well because it combines properties that suit continuous at-home use: non-invasive sampling, fast results, functional rather than purely compositional information, low per-measurement cost, and the possibility of genuine continuity rather than periodic snapshots.

Substantial questions remain open. How microbial communities recover after antibiotic exposure is incompletely understood, and recovery varies enormously between individuals. The specific mechanisms of the microbiota-gut-brain axis are still being worked out. The relative importance of individual microbial metabolites to human health remains debated. These are opportunities, and also reasons for humility about conclusions drawn today.

What the Evidence Actually Supports

The evidence reasonably supports the following:

• Microbiome composition correlates with digestive health status and with certain disease states, across 28 standardized case-control datasets covering ten conditions.
• Dietary changes can meaningfully shift microbiome composition within days.
• Individual postprandial responses to identical foods vary substantially, and microbiome features improve prediction of those responses.
• Fecal microbiota transplantation and standardized live biotherapeutics reduce recurrence of Clostridioides difficile infection in randomized trials.
• Greater microbiome diversity is generally, though not universally, associated with better outcomes.

The evidence is considerably less clear on:

• What an optimal microbiome composition looks like for any given individual.
• Whether current probiotic or supplement approaches benefit otherwise healthy people, given that mucosal colonization is person-specific and often absent.
• The precise causal mechanisms linking specific microbes to specific diseases, since roughly half of disease-associated genera respond to more than one disease.
• Whether correcting dysbiosis without addressing its underlying cause produces durable benefit.

That last point deserves emphasis. An observational finding that depleted butyrate producers are associated with a digestive condition does not establish that restoring them resolves it, because the cause of the depletion may itself need addressing.

The Convergence Point

The convergence of multi-omics approaches, machine learning, miniaturized analytical systems and rigorous clinical validation is creating real opportunities for precision medicine in digestive health. VOC analysis is a meaningful component of that convergence, enabling non-invasive monitoring of microbial ecosystem function.

SNIFR's technology is in development and has not been clinically validated. It is designed to surface patterns in gut health and, as a longer-term design goal, to work toward flare-up prediction as an early warning system. It is not designed or claimed to detect, diagnose or predict disease. Claims that outrun evidence, in this field more than most, undermine both public trust and scientific credibility.

Frequently Asked Questions

What is the latest gut microbiome research telling us?

The most consequential shift is from cataloguing which bacteria are present to measuring what they are doing. Reference gene catalogues have grown from 3.3 million genes in 2010 to 9.9 million in 2014, and integrated projects now combine metagenomics, metatranscriptomics, metabolomics and volatile organic compound analysis. That shift is what makes continuous, non-invasive monitoring plausible.

Is personalized nutrition based on the microbiome actually supported by evidence?

Partly. Zeevi and colleagues monitored 800 people across 46,898 meals and showed glycemic responses to identical foods vary widely and that microbiome features improve prediction, a result replicated in the 1,002-person PREDICT 1 study. What remains less settled is how reliably any commercial service can translate a stool sample into an optimal individual diet.

How is VOC analysis different from microbiome sequencing?

Sequencing produces a compositional snapshot of which organisms are present, while VOC analysis measures functional metabolic output in near real time. The same species can produce different volatile profiles depending on what it is fermenting. That is why the two approaches are complementary rather than interchangeable.

Why does machine learning matter in gut microbiome research?

Modern microbiome datasets contain far more variables than traditional statistical methods were designed to handle. Machine learning can find structure in that dimensionality. The important caveats are external validation, since models trained on one cohort lose accuracy on another, and explainability, since clinicians will not act on an opaque output.

Can gut microbiome monitoring predict disease before symptoms appear?

Prediction before symptoms is the goal of the field, not an established capability. Observational work has linked microbiome patterns to healthy ageing and survival in more than 9,000 individuals, but prediction windows for specific digestive conditions are not clinically validated. SNIFR's flare-up prediction work is a design goal, not a proven function.

Why do different laboratories give different microbiome results?

Because protocol choices matter as much as biology. The Microbiome Quality Control consortium sent identical blinded fecal specimens to 15 laboratories and found that DNA extraction protocol and bioinformatic pipeline introduced variation comparable to genuine differences between people. This is the main obstacle to comparing results across services and studies.

Do probiotics actually colonize the gut?

Not reliably, and not in everyone. Using endoscopic mucosal sampling rather than stool, Zmora and colleagues found colonization by an eleven-strain probiotic was person-specific, with some people resisting it entirely, and that stool shedding did not indicate whether colonization had occurred. A companion study found probiotics delayed native microbiome recovery after antibiotics.

References

• Qin J, Li R, Raes J, et al. A human gut microbial gene catalogue established by metagenomic sequencing. Nature. 2010;464(7285):59-65. doi:10.1038/nature08821
• Li J, Jia H, Cai X, et al. An integrated catalog of reference genes in the human gut microbiome. Nature Biotechnology. 2014;32(8):834-841. doi:10.1038/nbt.2942
• The Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature. 2012;486(7402):207-214. doi:10.1038/nature11234
• Proctor LM, Creasy HH, Fettweis JM, et al. The Integrative Human Microbiome Project. Nature. 2019;569(7758):641-648. doi:10.1038/s41586-019-1238-8
• Lloyd-Price J, Arze C, Ananthakrishnan AN, et al. Multi-omics of the gut microbial ecosystem in inflammatory bowel diseases. Nature. 2019;569(7758):655-662. doi:10.1038/s41586-019-1237-9
• Zeevi D, Korem T, Zmora N, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell. 2015;163(5):1079-1094. doi:10.1016/j.cell.2015.11.001
• Berry SE, Valdes AM, Drew DA, et al. Human postprandial responses to food and potential for precision nutrition. Nature Medicine. 2020;26(6):964-973. doi:10.1038/s41591-020-0934-0
• Asnicar F, Berry SE, Valdes AM, et al. Microbiome connections with host metabolism and habitual diet from 1,098 deeply phenotyped individuals. Nature Medicine. 2021;27(2):321-332. doi:10.1038/s41591-020-01183-8
• Krishnamoorthy A, Chandrapalan S, Ahmed M, et al. The diagnostic utility of volatile organic compounds in inflammatory bowel disease: a systematic review and meta-analysis. Journal of Crohn's and Colitis. 2024;18(2):320-330. doi:10.1093/ecco-jcc/jjad132
• Zheng J, Sun Q, Zhang M, et al. Noninvasive, microbiome-based diagnosis of inflammatory bowel disease. Nature Medicine. 2024;30(12):3555-3567. doi:10.1038/s41591-024-03280-4
• Wirbel J, Pyl PT, Kartal E, et al. Meta-analysis of fecal metagenomes reveals global microbial signatures that are specific for colorectal cancer. Nature Medicine. 2019;25(4):679-689. doi:10.1038/s41591-019-0406-6
• Duvallet C, Gibbons SM, Gurry T, et al. Meta-analysis of gut microbiome studies identifies disease-specific and shared responses. Nature Communications. 2017;8(1):1784. doi:10.1038/s41467-017-01973-8
• Topcuoglu BD, Lesniak NA, Ruffin MT, et al. A framework for effective application of machine learning to microbiome-based classification problems. mBio. 2020;11(3):e00434-20. doi:10.1128/mBio.00434-20
• Przymus P, Rykaczewski K, Martin-Segura A, et al. Deep learning in microbiome analysis: a comprehensive review of neural network models. Frontiers in Microbiology. 2025;15:1516667. doi:10.3389/fmicb.2024.1516667
• Li P, Luo H, Ji B, et al. Machine learning for data integration in human gut microbiome. Microbial Cell Factories. 2022;21(1):241. doi:10.1186/s12934-022-01973-4
• Sinha R, Abu-Ali G, Vogtmann E, et al. Assessment of variation in microbial community amplicon sequencing by the Microbiome Quality Control (MBQC) project consortium. Nature Biotechnology. 2017;35(11):1077-1086. doi:10.1038/nbt.3981
• Zmora N, Zilberman-Schapira G, Suez J, et al. Personalized Gut Mucosal Colonization Resistance to Empiric Probiotics Is Associated with Unique Host and Microbiome Features. Cell. 2018;174(6):1388-1405.e21. doi:10.1016/j.cell.2018.08.041
• Suez J, Zmora N, Zilberman-Schapira G, et al. Post-Antibiotic Gut Mucosal Microbiome Reconstitution Is Impaired by Probiotics and Improved by Autologous FMT. Cell. 2018;174(6):1406-1423.e16. doi:10.1016/j.cell.2018.08.047
• Kelly CR, Khoruts A, Staley C, et al. Effect of fecal microbiota transplantation on recurrence in multiply recurrent Clostridium difficile infection: a randomized trial. Annals of Internal Medicine. 2016;165(9):609-616. doi:10.7326/M16-0271
• Feuerstadt P, Louie TJ, Lashner B, et al. SER-109, an Oral Microbiome Therapy for Recurrent Clostridioides difficile Infection. New England Journal of Medicine. 2022;386(3):220-229. doi:10.1056/NEJMoa2106516
• Wilmanski T, Diener C, Rappaport N, et al. Gut microbiome pattern reflects healthy ageing and predicts survival in humans. Nature Metabolism. 2021;3(2):274-286. doi:10.1038/s42255-021-00348-0
• Valdes AM, Walter J, Segal E, et al. Role of the gut microbiota in nutrition and health. BMJ. 2018;361:k2179. doi:10.1136/bmj.k2179

SNIFR is designed to provide insights about gut health patterns, not to diagnose or treat medical conditions. Individual results may vary as gut health is influenced by numerous factors including diet, stress, sleep, and genetics. SNIFR is currently in development, and features described may evolve before commercial release.