My UM
My UM
Search

Speaking the Same Language: Building a Stronger Foundation for Microbiome Research

If you are reading this post, chances are that you work in microbiome research or that you are interested in. It is a relatively new field, especially when compared to other fields such as microbiology, ecology, or life-sciences. Nonetheless, in just a few decades, it evolved from a niche novelty to a central player in our understanding of human health, disease, and the environment. With such growth, though, comes a challenge: the faster a field moves, the more important it becomes to make sure everyone is moving in the same direction. That's where standardization comes in, and it's a conversation very much active within the microbiome community. A great example of it is the collective editorial “A blueprint for contemporary studies of microbiomes” published in the April 2025 issue of the journal Microbiome [1]. The editorial was co-authored by over twenty researchers from institutions across four continents, including senior editors of the journal itself. It's a piece of academic writing worth reading carefully, because the standards it calls for reflect real, well-documented challenges that have been hindering the field for years.

In this post we will review some of the most important points touched in the editorial.

Shared_language

A Shared Language Needs a Shared Vocabulary

Language is an excellent place to start. All fields of science depend on some form of shared vocabulary, but the microbiome field hasn’t yet settled on one. The terms microbiota and microbiome, for instance, are still often used as synonyms, but they refer to different things: one is the community of microorganisms itself, the other encompasses that community together with its metabolites, structural elements, and surrounding environment [1]. Similarly, "16S rRNA gene amplicon sequencing" and "metagenomics" describe fundamentally different approaches, the first profiles who is there using a single marker gene, the second sequences everything in the sample and can tell you what the community is actually capable of doing, and keeping them distinct matters for how we interpret and compare results [1, 2].

Then there's "abundance." When a study reports amplicon or metagenomic data, it is almost always reporting relative abundances (i.e., proportions, percentages, parts per milion, etc.) that only makes sense in relation to something (or everything) else in the sample. Sequencing can’t say how many bacteria are there; it can say who is there relative to whom. Without the qualifier "relative," the term quietly implies an absolute amount or count that the data simply don't contain [1]. This isn't just a matter of precision in writing; analyses designed for absolute count data can produce systematically different, or even outright mathematically meaningless, results when applied to compositional data, a point we'll return to below.

 

Experimental design: The Decisions That Affect Everything Downstream

Many microbiome studies have been underpowered by conventional metrics, making it harder to distinguish true biological effects from noise, and harder still to replicate results across labs [2]. Cross-sectional study designs, a snapshot in time, are still the norm, even though longitudinal designs are better suited to establishing whether microbiome differences precede or follow a health outcome. Standardizing the collection of key metadata, diet, medication use, age, BMI, stool consistency, and agreeing on which confounders to account for would go a long way toward making studies more comparable [2, 5].

 

Sample Collection, Storage, and DNA Extraction

These are a source of variability that is increasingly well-characterized, even if not yet fully solved. Different extraction kits yield systematically different results, particularly for Gram-positive bacteria that require mechanical cell disruption to lyse effectively [1, 2]. Low-biomass samples such as, blood, lung tissue, and biopsies, present a specific challenge: DNA extraction reagents themselves contain trace microbial DNA (sometimes called the "kitome"), which can become a meaningful fraction of the signal when the biological sample contains very little DNA to begin with [4]. A 2019 survey of published studies found that the inclusion of negative and positive controls, tools that help researchers distinguish biological signal from technical noise, was less common than it ideally should be [4]. The field is moving toward making these controls standard, and that's an encouraging direction.

 

Controls Are “Must Have” Standards

The case for controls is straightforward and yet still often overlooked. Negative controls (blanks that travel through the full extraction and sequencing pipeline alongside real samples) show what's coming from reagents and the environment. Positive controls, mock communities of known composition, let you verify that the pipeline is detecting what it should and how it should [3, 4]. For low-biomass studies in particular, where contamination, well-to-well leakage between samples, and batch-to-batch processing differences can generate signals that are difficult to distinguish from noise, these controls are not just good practice, they're essential for drawing defensible conclusions [8].

 

Bioinformatics: Open Tools, Versioned Pipelines, Reproducible Code

Bioinformatics pipelines have also matured considerably, with a clear shift from Operational Taxonomic Units (OTUs) clustering toward Amplicon Sequence Variants (ASVs): exact sequence representations that are more reproducible and allow direct comparison across datasets without re-clustering [2, 5]. Alongside this, there is growing agreement around the need to report software and packages versions, make analysis code publicly available, and specify the name and version of the taxonomic database used, details that are necessary for others to reproduce or build on published work [1]. Formal benchmarking of new methods, using mock communities and cross-validation approaches, is increasingly supported by initiatives like CAMI (Critical Assessment of Metagenome Interpretation) [3].

 

“Microbiome Datasets Are Compositional: And This Is Not Optional” [6]

I took the title of this section directly from a keystone article of 2017 to emphasise its message. But why is this such a big problem? Since relative abundances must sum to a constant, an apparent increase in one taxon is mathematically linked to apparent decreases in others, regardless of what is actually happening. This is compositionality in a nutshell. Compositionally aware approaches, like Centred-Log-Ratio data transformation or tools like ALDEx2 or ANCOM-BC, among others, are designed with compositionality in mind [1, 2].

That said, depending on the question and the analysis, accounting for compositionality is not always necessary or even advisable. This is what makes it genuinely tricky: whoever is planning or performing the statistical analysis needs to understand the data well enough to know when compositionally-safe methods are required. And there is no single best option or best tool as each comes with their own trade-offs and assumptions. Rather, there are only best practices, which is precisely why statistical planning should happen from the very beginning of a study.[1, 2]. Alpha-diversity metrics also need careful handling: some commonly used estimators (Chao1, ACE) are not appropriate for ASV-based data, and the Shannon index, while widely used, is not linearly proportional to diversity in a way that makes "significant" differences easy to interpret [1]. More intuitive alternatives, like the effective number of species (i.e., Effective Shannon), are worth adopting more broadly.

 

Metadata Standards 

Finally, metadata and data sharing. A 2022 analysis of publicly deposited microbiome datasets found that consistent metadata, beyond project name and collection date, was largely absent [7]. This limits the reuse value of data that, in many cases, took years to generate. The MIxS framework provides a minimum standard for what metadata should accompany sequencing data, and the FAIR principles offer a broader philosophy for how scientific data should be managed and shared [7]. Reporting checklists like STORMS (for human microbiome studies) and the forthcoming STREAMS guidelines (for environmental and non-human studies) make these principles practical [1].

 

Conclusions

The tools to address all of this are largely in place. What the field benefits most from now is a collective commitment to use them consistently, across labs, journals, and disciplines. Standardization is not a constraint on scientific creativity, but rather is what allows individual discoveries to build on what others did before and collect into reliable knowledge. 

It is, quite simply, what makes it possible to stand on the shoulders of giants.

 

References

  1. Bindels LB, et al. Microbiome 2025;13:95.

  2. Knight R, et al. Nat Rev Microbiol 2018;16:410–422.

  3. Bokulich NA, et al. Comput Struct Biotechnol J 2020;18:4048–4062.

  4. Hornung BVH, et al. FEMS Microbiol Ecol 2019;95:fiz045.

  5. Sorbie A, et al. iScience 2022;25:103998.

  6. Egozcue JJ, et al.  Front Microbiol 2017: 8:2224.

  7. Cernava T, et al. Environ Microbiome 2022;17:33.

  8. Austin GI & Korem T. J Infect Dis 2024;233(Suppl 1):76–86.