miaSim: an R/Bioconductor package to easily simulate microbial community dynamics

Abstract

Abstract Microbiomes never stop changing. Their compositions and functions are shaped by the complex interplay of intrinsic and extrinsic drivers, such as growth and migration rates, species interactions, available nutrients and environmental conditions. Mathematical models help us make sense of these complex drivers and intuitively explain how, why and when specific microbiome states are reached while others are not. To make simulations of microbiome dynamics intuitive and accessible, we present miaSim. miaSim provides users with a wide range of possibilities to match their specific assumptions and scenarios, starting from a core implementation of four widely used models (namely the stochastic logistic model, MacArthur's consumer‐resource model, Hubbell's neutral model and the generalized Lotka‐Volterra model) and several of their derivations. The diverse model implementations share the same data structures and, whenever possible, share state variables, which significantly facilitates cross‐model combinations and comparisons. We combined and simulated some published examples of microbiome models in miaSim and performed cross‐model comparisons and tested diverse model assumptions. Our examples illustrate the reliability, robustness and user‐friendliness of the package. In addition, miaSim is accompanied by miaSimShiny, which allows users to explore the parameter space of their models in real‐time in an intuitive graphical interface. miaSim is fully compatible with the ‘miaverse’, an R/Bioconductor framework for microbiome data science, allowing users to combine and compare model simulations with microbiome datasets. The stable version of miaSim is available through Bioconductor 3.15, and the version for future development is available at https://github.com/microbiome/miaSim. miaSimShiny is available at https://github.com/gaoyu19920914/miaSimShiny. We anticipate that miaSim will significantly facilitate the task of simulating microbiome dynamics, highlighting the role of ecological simulations as important tools in microbiome data science.