How sample heterogeneity can obscure the signal of microbial interactions

Abstract

Compositional data are commonly subjected to computational tools such as correlation networks, null models, and dynamic models, with the goal of identifying the ecological processes structuring microbial communities. Researchers applying these methods assume that the signs and magnitudes of species interactions and vital rates can be reliably parsed from compositional data. However, we contend that this assumption is violated when sample units contain any underlying spatial structure. Here, we show how three phenomena -- Simpsons paradox, context--dependence, and nonlinear averaging -- can lead to erroneous conclusions about population parameters and species interactions when samples contain heterogeneous mixtures of populations or communities. At the root of this issue is the fundamental mismatch between the spatial scales of species interactions (micrometres) and that of a typical marker gene sample (millimetres to centimetres). These issues can be overcome by measuring and accounting for spatial heterogeneity at very small scales, which will lead to more reliable inference of the ecological mechanisms structuring natural microbial communities.