Tracking strains predicts personal microbiomes and reveals recent adaptive evolution

An individual’s microbiome consists of a diverse set of bacterial strains that encode rich information on its colonization and evolutionary history. Here, we introduce a versatile and straightforward reference-based strain tracking approach (StrainTrack) that determines whether distinct metagenomes carry closely-related strains based on gene presence and absence profiles. We show that StrainTrack can predict whether two metagenomes originate from the same donor via counting the number of species sharing closely-related strains, achieving >96% specificity, and ∼100% sensitivity. When applied to the metagenomes of adult twins in the TwinsUK registry, we identify six cases of closely-related strains carried by both twins, potentially over decades of colonization.