The core microbiome as a reproducible, data-driven abstraction, not a biological entity

Abstract

Microbial communities are highly diverse, making it difficult to distinguish stable ecological patterns from stochastic variation. The core microbiome offers a widely used framework for simplification, but prevalence-based definitions are often criticized for relying on arbitrary thresholds. We propose an alternative view in which the core microbiome is not a biologically complete or functionally exhaustive subset, but a data-driven abstraction that preserves dominant ecological information under strong dimensional reduction. Using bacterial (16S rRNA) and fungal (ITS2) communities from leaves, roots, and rhizosphere of clonally replicated pedunculate oak (Quercus robur) across a continental environmental gradient, we defined cores using a non-arbitrary threshold derived from the prevalence distribution. Despite retaining fewer than 6% of bacterial and 3% of fungal OTUs, cores preserved composition patterns, reproduced site differentiation, retained most predicted functional information, maintained network structure, and were robust across a wide range of sampling efforts. Prevalence and abundance captured distinct ecological dimensions: consistently occurring taxa were not necessarily the most abundant, and vice versa. Our results show that core value therefore lies not in identifying the most important microorganisms, but in providing a reproducible, information -preserving representation of complex communities - though core taxa nonetheless retained fundamental ecological and functional roles.