Microbiomics risks being drowned in a tsunami of its own hype. Jonathan Eisen, a microbiologist and blogger at the University of California, Davis, bestows awards for “overselling the microbiome”; he finds no shortage of worthy candidates.

Previous 'omics' fields have faltered after murky work slowed progress2. Technological advances that allowed researchers to catalogue proteins, metabolites, genetic variants and gene activity led to a spate of associations between molecular states and health conditions. But painstaking further work dampened early excitement. Most initial connections were found to be spurious or, at best, more complicated than originally believed.

The history of science is replete with examples of exciting new fields that promised a gold rush of medicines and health insights but required scepticism and years of slogging to deliver even partially. As such, the criteria for robust microbiome science are instructive for all researchers. As excitement over the microbiome has filtered beyond academic circles, the potential mischief wrought by misunderstanding encompasses journalists, funding bodies and the public.

Crucial questions

Here are five questions that anyone conducting or evaluating this research should ask to keep from getting carried away by hype.

Can experiments detect differences that matter? Profiling a microbiome could produce a catalogue at the level of phyla, species or genes. Much work relies on analysis of 16S rRNA, an ancient gene that tolerates little variation and so is reliably found across the bacterial kingdom. But this allows only a coarse sorting. For example, microbiomes associated with obesity have been distinguished by different ratios of bacterial phyla, which encompass a staggering range of diversity. If this criterion were used to characterize animal communities, an aviary of 100 birds and 25 snails would be considered identical to an aquarium with 8 fish and 2 squid, because each has four times as many vertebrates as molluscs. Even within a single species, strains often differ greatly in the genes they contain.

Press officers must stop exaggerating results, and journalists must stop swallowing them whole. 

Modern technology now allows for finer distinctions: we can study more genes in a sample, an ability that may enable us to decipher 'metabolic networks' revealing the biochemical reactions that a microbiome can perform. This kind of analysis could identify gene combinations, potentially from multiple species across a microbial community, that affect health for good or ill. However, pinning an outcome to any particular entity is likely to be hard unless the networks are already well characterized.

To take a simple example from a single bacterial species, we could show that vaccination eliminated 30% of known pneumococcal strains in a human population — but only because we knew in advance to focus on the genes targeted by the vaccine3. Our ability to identify functional differences in closely related genes is rarely sophisticated enough to pull out important genes or networks if we do not know what to look for in the first place. Moreover, genomes are littered with clues both true and false, such as 'hypothetical proteins' and genes that are understood poorly or not at all, but could make for important differences in what metabolic networks do.

We need to be able to identify functional differences in closely related genes from sequence alone. Until then, we must remember that apparent similarities might cloak important differences.

Does the study show causation or just correlation? A separate question is raised when distinct microbiomes can be identified and associated with diseases or other conditions. Then we are left with the chestnut of causes and correlates. Sometimes, a particular microbiome found in association with disease will be merely a bystander4.

A 2012 article comparing the gut microbiomes of old people living in care homes with those of old people living in the community found distinct microbiomes that correlated with multiple scores of frailty5. After accounting for some potentially confounding factors, the authors proposed a causal relationship: diet altered the microbiome, which in turn altered health. This explanation fits the data, but the reverse causality — the potential for poor health to alter the gut microbiome — was not explored. Frailer people probably have less active immune systems and differences in digestion (such as the time required for food to pass through the stomach and intestines) — factors that could change the microbiome. This work is not the only example of this sort of confusion.

What is the mechanism? All scientists are taught the catechism that correlation is not causation, but correlation almost always implies some sort of causal relationship. We just don't know what it is. We must determine it with careful experiments.

In the past three or four years, studies have advanced from characterizing a broad community of mainly unculturable microbes to identifying functional elements, individual taxa or particular properties. We can now design experiments to precisely define actions of components of the microbiome6, for example by reconstituting communities but leaving out specific taxa, or by precisely measuring the biochemical activity of an experimental microbiome in an 'organ on a chip'7. A return to a reductionist approach is essential if we are to pinpoint both whether the microbiome affects human health, and exactly how it does so.

How much do experiments reflect reality? Even if the microbiome can have an experimental effect, it may not be an important cause of the symptoms seen in ill people.

Much work has addressed the role that gut flora have in obesity, and several studies have found associations between the gut microbiome and weight gain8. To assess whether this association was cause or consequence, researchers collected gut-microbiome samples from human twins (one obese, one not) and introduced the microbiota to mice. Mice previously colonized with an 'obese' microbiome lost weight when supplied with a 'lean microbiome', but only if also fed a normal or low-fat diet. Diet alone had little effect9. Although this elegantly controlled experiment suggests great potential for the microbiome and related therapies to affect health, it also shows the microbiome's limits: the effect was dependent on other factors, in this case diet.

Microbiome studies often rely on germ-free mice. These animals allow researchers to readily introduce an experimental microbiota. But they do not represent the animals' natural state and are typically unhealthy owing to the lack of a microbiome. So results may not predict responses in animals with flourishing microbiomes. Mice and their microbiomes are also adapted to a rather different niche from humans, so results may not be generalizable.

Could anything else explain the results? There are good reasons to think that bacteria influence us in a host of ways. But there are many other — possibly more important — influences, such as diet in the earlier example. Whenever a study links a microbiome to a disease, wise critics should ask whether other contributors to disease are considered, compared and reported.

The hype surrounding microbiome research is dangerous, for individuals who might make ill-informed decisions, and for the scientific enterprise, which needs to develop better experimental methods to generate hypotheses and evaluate conclusions. Funding agencies must not let their priorities be distorted by the buzz around the field, but look dispassionately at the data. Press officers must stop exaggerating results, and journalists must stop swallowing them whole. In pre-scientific times when something happened that people did not understand, they blamed it on spirits. We must resist the urge to transform our microbial passengers into modern-day phantoms.

The microbiota is booming — the number of studies mentioning ‘microbiome’ or ‘microbiota’ in their title or abstract grew from 11 in 1980 to over 13,000 in 2018. Almost any human disease you can think of has proposed links with the microbiome: inflammatory bowel disease, cancer, diabetes, obesity, atherosclerosis, fatty liver disease, malnutrition, autism, Alzheimer disease, depression, autoimmunity, asthma and on goes the list. Do you want to run faster, sleep better, live longer, or have more friends? There is bound to be a claim for a ‘microbiota fix’. An array of microbiome-based therapeutics are on the market or in development, ranging from prebioctics, probiotics, postbiotics, poop pills and biologics that target the gut microbiota to probiotic skin care and vaginal microbiota transplants. Furthermore, various companies offer testing of your (or your dog’s) gut microbiome and then make personalized diet and lifestyle recommendations and some even sell supplements. And for the so inclined there are now even smart toilets that send faecal bacterial counts directly to your smartphone. As a microbiologist, one cannot help but think of some of these developments as the Wild West of microbiome science.

However, it is clear that we live in close association with our microbiota and their functions are manifold. Furthermore, many human diseases are complex and for some, current treatments or preventative measures are insufficient and targeting the microbiota could be a potential way to deal with some of these challenges. The prime example is Clostridioides difficile infection, which can be efficiently treated with faecal microbiota transplantation. Owing to technological and analytical advances, our knowledge of microbiota diversity and functions and their roles in disease is rapidly increasing, and efforts are under way to go beyond disease associations and to translate this knowledge into therapeutics and applications. In large part, successful translation relies on tractable methods and systems that enable systematic and controlled testing of hypotheses and interventions. To highlight the latest developments in microbiome tractability and translation, we have put together a special Focus issue on these topics.

successful translation relies on tractable methods and systems that enable systematic and controlled testing

Models are essential tools to understand disease mechanisms and the specific molecular links that determine host–microbiota associations and interactions. In recent years, microbiome research has increasingly expanded to the use of simple animal models such as flies, zebrafish and worms. Angela Douglas argues that these simple models are cost-effective and time-efficient, and are particularly useful for large screens or multivariate experimental designs.

Once such a screen has identified a trait or relationship to target, microbiome engineering can be used to develop interventions. Chris Lawson, Trina McMahon and colleagues detail the design-build-test-learn cycle to efficiently engineer microbiomes for various applications. This approach has been successfully used in other fields such as metabolic engineering and provides a tractable path for microbiome translation.

One potential application is manipulating the human gut microbiota to treat or prevent diseases such as diabetes mellitus or cardiovascular disease. There is a clear link between nutrition, the gut microbiome and health. However, large individual differences make it difficult to predict the direction and strength of responses and thus to therapeutically manipulate this link. Eran Elinav and colleagues discuss the considerable variation in how individuals and their microbiome react to diet and how personalisation could optimize any therapeutic or preventative interventions that target this axis.

The human gut is a complex environment and Jeroen Raes and colleagues argue that a systems ecology perspective is needed to understand the gut microbiota. They provide an outline of how culturing and in vitro interaction experiments, combined with computer modelling, can help us to develop a functional and mechanistic understanding of the gut microbiota, which is the basis for successful translational projects.

In summary, advances in modelling, be it animal, in vitro or in vivo models, and ever increasing datasets raise great hope for microbiota translation and applications. In this context, a little bit of hype might not only have negative consequences. After all, the strong interest in microbiome science has led to funding initiatives and industry engagement (as reflected by predicted growth rates of the microbiome market of over 20% per year). It is time to further strengthen the scientific basis from which microbiome translation can grow.