It is now established that the microbial communities that inhabit humans (the microbiome) play an integral role in health and disease. Yet, much of the true diversity of these microbial systems are either poorly characterised or largely unexplored. Metagenomics offers the opportunity to characterise this diversity including at a sub-species (strain-level) resolution. I will present the challenges, and advancements in computational approaches to exploit this potential. In particular, I will exemplify their use to uncover functional strain-level features pertinent to disease and identify and track individual strains. I will also demonstrate that publicly available metagenomes can be utilised to explore the genetic and population structures of common but poorly characterised members of the microbiome. Focussing on Prevotella copri, for which there exists only a few isolate genomes, I will demonstrate how metagenomics has provided new insights into this enigmatic gut microbe. I will show that P. copri is not a monotypic species as previously assumed but rather four distinct sub-types. All four types have been a long-standing feature of the human ancestral microbiome, pre-dating human migration out of Africa, but have been rapidly lost in Westernised populations as a consequence of industrialisation over the past few hundred years. Thus, I will demonstrate that metagenomics is not only of epidemiological and clinical relevance but also a tool for asking fundamental questions regarding the true taxonomic diversity of these complex microbial communities.