The human microbiome is key to health and disease. Yet, mapping and understanding microbiome ecology and microbiome-host interactions is a challenge due to the complex genetic, metabolic and immunological interplay between microbial species and the human host. In this presentation, we explore various computational roads to a better mapping of microbial metabolism via constraint-based reconstruction and analysis. We will demonstrate how digital twins of personalised microbial communities can be generated and interrogated via in silico experimentation, and how deterministic modelling can be integrated with statistical modelling approaches for better contextualisation of omics results and even causal inference in the presence of unmeasured confounders. We conceptualise important attributes such as functional redundancy and metabolic equivalence to enable a more in-depth functional characterisation of microbial communities beyond pathway enrichment and gene annotations. In the end, we will then take a critical look at the introduced modelling paradigms, highlighting challenges, limitations, and potentials, in particular in the context of human cohort analysis.