New physics-based model shows healthy guts resist microbial chaos

The human gut is a vast ecosystem of microbes whose balance is central to health. When this balance, known as eubiosis, is disturbed, the result is dysbiosis, and is linked to conditions such as inflammatory bowel disease. Traditional models of microbial communities tend to focus on how many species are present and how abundant they are, but they rarely capture how these species interact. Yet the web of microbial interactions defines whether a community is stable or prone to collapse.

In a new study in eLife, researchers from the University of Padova in Italy, ETH Zurich in Switzerland, and Paris Cité University in France sought to fill this gap by using the tools of statistical physics to study microbial ecosystems. Specifically, they borrowed ideas from the theory of disordered systems, which was originally developed to understand complex materials like spin glasses. Their goal was to connect measurable data from metagenomic sequencing to theoretical models that describe how thousands of species interact. By doing so, they hoped to identify mathematical fingerprints that could distinguish healthy from unhealthy gut microbiomes.

The researchers used the disordered generalised Lotka-Volterra model (dgLV) — a framework that describes how populations of species change over time based on random interactions. Each microbial species’ growth rate depends on its own capacity to thrive and on the positive or negative effects of other species. The interaction strengths, which are impossible to measure directly, were treated as random variables.

Using cross-sectional gut microbiome data from 91 healthy and 202 diseased samples, they extracted a set of numbers that they used to summarise each community’s behaviour. Then, an algorithm adjusted the dgLV parameters so that the model’s predictions matched the empirical data. This allowed the researchers to infer the statistical structure of inter-species interactions and to quantify how ecological stability differed between the healthy and diseased samples.

The analysis revealed that healthy and diseased microbiomes occupy distinct zones in the model’s parameter space. That is to say, healthy communities had stronger but more heterogenous interactions, reflecting a balanced network where fluctuations are absorbed without destabilising the system. Diseased microbiomes had lower interaction diversity and higher randomness, indicating instability and loss of resilience.

Mathematically, the diseased systems were found to lie closer to a critical line that marks the transition to chaotic or unstable dynamics. This suggests that while healthy microbiomes operate in a stable equilibrium, diseased ones hover near disorder.

The new framework reframes gut health as a problem of statistical structure rather than a matter of simple composition. It implies that dysbiosis is less about which microbes are missing and more about how their interactions have weakened or become erratic. As a result, the study suggests that by examining collective behaviour, one might predict or even correct unstable microbial states.

The authors also acknowledged some limitations in their effort. For one, their model assumes a static set of interactions while real microbes evolve and respond to changing environments in different ways. They suggest future studies might allow interactions to fluctuate, capturing their dynamics more realistically.