M.C.C. One strength of microbiome research today has been provided by the development of high-throughput sequencing and biocomputational tools, which have enabled comprehensive analysis of microbial communities. The advances in sequencing technologies and increasing number of human metagenomics catalogues (the majority of which are gut and oral microbiomes) covering distinct human body sites, across populations and human conditions, are providing pivotal information about host–microorganism interactions and health outcomes. The development and implementation of deep learning and artificial intelligence tools in the microbiome field will provide novel information and insights — for example, the discovery of new antimicrobial peptides17 and antibiotic molecules18 – that can lead to potential therapies and diagnostics.

There are several important current limitations of microbiome research. The main limitations relate to the complexity and variability of the microbiome, as current studies are not large enough to encompass the entire microbial inter-variability and intra-variability (which depend on host factors, environmental factors and others). This variability has led to various initiatives to develop and implement standardized protocols or guidelines for analysis, biomarker discovery, and the development of targeted or precision therapies. It is also important to mention the limitations in studying low-biomass microbial environments mainly due to technical issues (detection limits), potential contamination, low DNA yields, amplification bias and difficulties in data interpretation. Cutting-edge, high-sensitivity, high-coverage and low-cost sequencing technologies along with robust bioinformatics tools are necessary to overcome these challenges. In addition, limited and curated reference databases are available and this limits the ability to accurately identify and characterize microorganisms. This limitation is most pronounced in non-bacterial microbiomes (such as fungi, viruses and archaea). Another limitation is related to the difficulty of assessing causality, leaving most studies with associations and correlations. It is very difficult to distinguish causal relationships from mere associations between microorganisms and health conditions. Available data provide correlations and associations but the specific mechanisms by which the microbiome influences human health are not fully understood. In addition, preclinical data with animal models to understand the mechanisms of action cannot be translated to humans or into effective clinical applications.

In addition, a limited number of long-term studies hinder our understanding of how changes in the microbiome affect health across the lifespan. We also have an incomplete understanding of the microbiota in early life and how it affects health outcomes in the short and long term. Moreover, many microbial species remain uncharacterized and even isolated, and their functions and interactions are not fully understood. Finally, regulatory aspects, for example, regarding use of probiotics, novel live biotherapeutics and FMT, are important. This includes the implications of the new European Union regulation on Substances of Human Origin, adopted in May 2024, which will impact microbiome research, and aims to create a more reliable and standardized framework for using microbiome-related substances in research and therapy across the EU.

Another relevant point relates to the use of microbiome data as, like genetics data, microbiome data is linked to the individual and contains specific information. Thus, potential concerns regarding the ownership, privacy and potential misuse of microbiome information need to be addressed.

S.D. I would say that the greatest strength of the microbiome field — our ability to generate prodigious amounts of data through sequencing — is also its greatest limitation. When sequencing platforms became cheaper and more accessible and computational tools for handling the enormity of the data were developed, the microbial world opened up in wonderful and exciting ways. It enabled us to bypass the existing challenges of cultivation and interrogate complex bacterial communities and their genomes in new ways. It also enabled us to study the microbiome composition of populations at scale. These were incredible advances that continue to generate considerable data. The downside is that we have become too reliant on these tools and we do not perform as many actual experiments as we should. We often do not follow up on findings from sequencing data with enough experiments to see whether we can falsify the hypotheses they generate. Perhaps most concerningly, we do not utilize our knowledge of basic microbiology, molecular biology and bacterial genetics enough as a means to understand and interrogate mechanisms. Recapitulating a phenotype through FMT is interesting but it is not a mechanism — we need to go deeper. This is part of the reason why development of targeted microbiome therapeutics has struggled. At the same time, I have seen some very clever and unique ideas and mechanisms proposed and explored commercially, so I remain optimistic and firmly believe that there is real potential in our field worth pursuing. We also have what I feel is a very open community of researchers who enjoy collaborating and sharing their work. When you go to microbiome meetings, the energy is positive and there is a genuine mutual interest in each other’s work, an open-mindedness, and a desire to do good through our discoveries — I have noticed this ever since I was a trainee.

T.S.G. The most challenging aspect of the gut microbiome is its variability, even among individuals without disease, which is influenced by a multitude of host-associated factors such as urbanization, diet, age and medications. As the definition of a ‘normal’ microbiome changes across different populations, the reference lists for disease-associated taxa, the health-associated ‘therapeutic’ taxa, as well as taxa that drive responses to different therapies are all expected to show cohort-specific variations. Integrated multi-cohort meta-analyses and multi-omics investigations linking the variations or alterations in the microbiome to other host-associated traits have tried to profile these variations.

However, the majority of gut microbiome studies have focused on industrialized population cohorts. Limited availability of microbiomes from non-industrialized populations from low-income and middle-income countries (LMICs) makes the robust identification of microbiome-derived markers of health and disease in these populations challenging. Prohibitive costs for performing deep metagenome sequencing and integrated multi-omics are going to be a major challenge for translatable microbiome research in LMICs.

Another challenge is the identification of microbiome-derived therapeutic and diagnostic signatures. The gut microbiome is a community that is in close homeostasis with the host as well as with itself. Thus, the association of a community member with host health and/or disease is in many cases observed to be context dependent, contingent on the state of the host as well as on the microbiome itself.

Using the faecal microbiome as a proxy for the gut, primarily due to the lack of availability of non-invasive sampling techniques that grant access to specific regions within the gut (for example, stomach or duodenum), is also a limitation. Another limitation has been the focus on species-level or genus-level composition of mostly or predominantly bacterial or archaeal lineages to represent the gut microbiome, while ignoring intra-species strain-specific variations as well as variations in other components of the microbiome such as fungi and viruses, which have been minimally explored due to the costs of high-depth metagenome sequencing.

Given these limitations of cohort-specific variations, context-dependent associations and limited availability from specific population groups, translatability of microbiome research is impeded by difficulties in obtaining regulatory approvals. Approvals and success of microbiome-derived therapeutics depend on a holistic understanding of the total microbiome of a patient. Future success of microbiome research towards clinical translation thereby depends on precise reporting and monitoring of personalized outcomes.