Abstract
Gut microbiota are essential for maintaining host health, for example by providing protection against pathogens. This has prompted numerous studies to explore the composition and diversity of microbial communities using metagenomics techniques. Although this allows some degree of insight, there remains a shortcoming in understanding the interactions and temporal dynamics of these communities in greater detail. Moreover, this knowledge gap becomes further pronounced as we also consider the impact of external perturbations, for example antibiotic treatment, on the long-term stability of these microbial communities. To address this, we developed a mechanistic modelling framework based on the generalised Lotka–Volterra model to predict microbial compositional changes over time and to assess its sensitivity to applied external perturbation. Essentially, whether or not the microbial community bounces back to its original configuration once the perturbation has stopped. Using in-silico data and publicly available data derived from 16S rRNA sequencing, we estimated microbial growth rates and their mutual interactions. This model relies on absolute abundance counts, which can be estimated from total microbial biomass measurements, and the data is organised into the topmost abundant taxa organised at the genus level to prevent over-fitting. Bayesian inference was used to estimate model parameters from these abundance counts. Perturbations were represented by imposing seasonal changes in microbial growth rates that fluctuated about their non-perturbed values. After fitting the model to these datasets, we explored different applications of the perturbation signal and evaluated its impact on the long-term stability of the community dynamics. The model shows that, even if the intensity of the perturbation is the same (e.g., a given dosage of antibiotics), there are specific frequencies at which the perturbation is administered that can cause pronounced responses (resonance). Essentially, we can induce large deviations in microbial abundances from their equilibrium values and even drive some taxa to extinction. Applications of this approach include identifying optimal antibiotic treatment regimens to minimise the emergence of superbugs resistant to antibiotics and informing personalised strategies for maintaining a healthy gut microbiota.