The Metagenome-Assembled Genome Inventory for Children (MAGIC)

Existing microbiota databases are biased towards adult samples, hampering accurate profiling of the infant gut microbiome. Here, we generated a Metagenome-Assembled Genome Inventory for Children (MAGIC) from a large collection of bulk and viral-like particle-enriched metagenomes from 0-7 years of age, encompassing 3,299 prokaryotic and 139,624 viral species-level genomes, 8.5% and 63.9% of which are novel. MAGIC improves early-life microbiome profiling, with the greatest improvement in read mapping observed in Africans. We then identified 54 candidate keystone species, including several Bifidobacterium spp. and four phages, forming guilds that fluctuated in abundance with time. Their abundances were reduced in preterm infants and were associated with childhood allergies. By analyzing the B. longum pangenome, we found evidence of phage-mediated evolution and quorum sensing-related ecological adaptation. Together, the MAGIC database recovers previously uncharted genomes that enable characterization of dynamics of early-life microbiomes, identification of candidate keystone species, and strain-level study of target species.