The latent simplicity of microbial ecological interactions

Microbial communities carry out functions of ecological, clinical and industrial importance. These functions arise from individual species contributions and from pairwise and higher-order interactions, whose number grows exponentially with community size. Yet community function is often surprisingly predictable. Here we show that this predictability does not require weak interactions. Instead, interactions across orders are strongly coordinated: those involving a given species are approximately proportional to related interactions one order below. This coordination causes species effects to vary together across community backgrounds, allowing function to be described by only one or two collective variables. We demonstrate this organization across dozens of combinatorial bacterial experiments and multiple environments. For total biomass, consumer-resource models and experiments identify the dominant collective variable with species yield, a trait that remains stable as abundances change through community dynamics. Microbial communities are therefore predictable not because their interactions are simple, but because their complexity is coherently organized. Our work provides a framework for understanding and ultimately designing community function from species traits.