Abstract
The assembly of microbial communities through sequential invasions of microbial species is challenging to study experimentally. Here, I used genome-scale metabolic models of multiple species to model community assembly. Each such model represents all known biochemical reactions that a species uses to build biomass from nutrients in the environment. Species interactions in such models emerge from first biochemical principles, either through competition for environmental nutrients, or through cross-feeding on metabolic by-products excreted by resident species. I used these models to study 250 community assembly sequences. In each such sequence, a community changes through successive species invasions. During the 250 assembly sequences, communities become more species-rich and invasion-resistant. Resistance against both constructive and destructive invasions - those that entail species extinction - is associated with high community productivity, high biomass, and low concentrations of unused carbon. Competition for nutrients outweighs the influence of cross-feeding on the growth rate of individual species. In a community assembly network of all communities that arise during the 250 assembly sequences, some communities occur more often than expected by chance. These include invasion resistant "attractor" communities with high biomass that arise late in community assembly and persist preferentially because of their invasion resistance. Genome-scale metabolic models can reveal generic properties of microbial communities that are independent of the resident species and the environment.