Fig. 1. Schematic overview of the modelling framework established in this study.

Metagenomic reads were mapped onto a reference set of AGORA¹⁸ genomes, and the strain-level relative abundances were retrieved. For each metagenomic sample, a personalised microbiome model was constructed. The models were parameterised further with a simulated “Average European” diet retrieved from the Virtual Metabolic Human¹⁷ database. The metabolic profile of each microbiome model was then computed with distributed flux balance analysis⁵³. Required input data are shown in green and required tools are shown in purple.