MOFA: microbial optimization without forced altruism

Microorganisms typically exist in communities, where interactions among them define the complexity of these ecosystems. Developing in silico frameworks to investigate the behavior and functionality of these communities is therefore essential for advancing our understanding of microbial ecology. In recent years, several computational modeling frameworks based on genome-scale models have been developed for the community-level analysis of microbial systems.

Here, we introduce microbial optimization without forced altruism (MOFA), a bilevel optimization framework that considers both species-level and community-level fitness criteria. By imposing constraints on species biomass in the outer problem, it prevents the forced altruism observed in previous algorithms. We applied MOFA to a toy model and to community models of Desulfovibrio vulgaris and Methanococcus maripaludis, which exhibit a cross-feeding relationship that causes the community objective to override individual fitness goals by prioritizing the export of metabolites for other community members. For this microbial community, a comparison with the results of NECom, OptCom, and Joint-FBA shows that MOFA yields predictions that better match the experimental results. Additionally, for pairs with a cross-feeding relationship in which exported metabolite production associated with this mutual interaction competes with species biomass, such as D. vulgaris and M. maripaludis, NECom fails to predict the community growth rate, whereas our method succeeds.

MOFA effectively analyzes community growth rates without relying on forced altruism. In cases where NECom fails to predict community growth, MOFA successfully predicts these growth rates. Furthermore, MOFA enhances computational efficiency by eliminating the need for the binary variables required in the NECom algorithm.