Trait-based approach is a powerful framework linking community structures and environments. In contrast to its success in plant ecology, trait-based description of prokaryotic community has been hampered by the limited availability of prokaryotic trait information, especially that of unculturable prokaryotes. Here, we propose habitat-based analysis, which classifies prokaryotic community members by their estimated habitats. The key idea is to systematically infer the habitat of each prokaryote using comparative metagenomics: by comparing soil metagenomes with other metagenomes, for example, soil-specific prokaryotes can be identified. This provides prokaryotic habitat information — a trait reflecting the prokaryotes’ physiology and function — without reliance on culture-based experiments, incorporating the otherwise neglected unculturable members into the framework of trait-based approach.
As an implementation of this idea, we constructed an “environmental atlas of prokaryotes,” or ProkAtlas, which comprehensively links 16S rRNA gene sequences to prokaryotic habitats. By projecting the prokaryotic community sequences to ProkAtlas, the “habitat composition” of the microbial community can be estimated. When applied to empirical datasets of soil, human gut, carcass, and aquatic microbiomes, our habitat-based analyses effectively highlighted environment-prokaryotic community structure relationships, processes of prokaryotic community succession, and spatial migration of prokaryotes.