Catalog of 65,000 vaginal microbial genomes reveals seven loci associated with host genetic variation
Background: Filling the reference-genome gap in vaginal microbiome research
A reference catalog containing 65,055 vaginal microbial genomes has been assembled. Using it, researchers identified seven human genetic loci associated with vaginal microbes and replicated the associations in separate cohorts. Published in Nature Genetics on June 11, 2026, the study examined both genetic diversity within the microbes and associations with host genetic variation.
The 16S ribosomal RNA (rRNA) gene analysis widely used in vaginal microbiome research is useful for classifying bacteria, but has limited resolution for distinguishing species- and strain-level differences and functions. It also does not cover nonbacterial microbes such as viruses and fungi. Shotgun metagenomic sequencing, which reads DNA from multiple organisms together, can broaden this scope, but requires sufficient reference genomes against which to compare sequences.
Existing vaginal microbial datasets have been weighted toward Western populations and pregnant participants, leaving the diversity of non-Western populations underrepresented. The fact that human DNA accounts for more than 90% of DNA in vaginal samples has also made microbial genome reconstruction difficult. The researchers aimed to address these gaps by integrating large-scale sample collections from China with international public datasets.
Key findings: A genome resource and replicated host genetic associations
The researchers integrated 10,665 metagenomes from China’s Peacock project and 2,967 public metagenomes, along with previously reconstructed genomes and data from cultured isolates. The resulting Global Vaginal Metagenome-assembled Genome Catalog (GVMG) comprises 36,059 prokaryotic genomes, 43 fungal genomes, and 28,953 viral genomes. The prokaryotes were classified into 890 species-level genome groups, 116 of which were absent from public reference databases.
Geographic differences in genomes and functions were also observed within the same microbial species. Lactobacillus iners showed more pronounced geographic genomic differentiation than Lactobacillus crispatus. This finding concerns genomic differences within the microbes themselves and should be distinguished from genetic variation in human hosts.
The analysis combining metagenomics and genome-wide association analysis (M-GWAS) used a Chinese discovery cohort of 3,137 participants and validation cohorts of 3,227 and 506 participants. In the discovery cohort, the researchers analyzed associations between about 5.46 million common human genetic variants and 54 microbial species, adjusting for age, sequencing read count, and host genetic population structure. The outcomes assessed were species presence or relative abundance.
Seven loci met the study-wide significance threshold and showed associations in the same direction in at least one validation cohort. In particular, associations of OPRK1 with Ureaplasma urealyticum, ADAP1 with Lactobacillus mulieris, and a variant near PRAMEF1 with Bifidobacterium piotii replicated in both validation cohorts.
Meaning and outlook: A basis for follow-up mechanistic studies
This work provides a reference resource for more precise identification of vaginal microbes and candidates for follow-up experiments. Because genomes and functions can differ even within a species, the findings offer a basis for considering within-species diversity when comparing microbial distributions across populations.
However, a statistical association between genetic variants and microbes does not establish a direct regulatory mechanism. The researchers also identified additional experiments to clarify molecular interactions as a next step. Unlike the catalog, which integrates international datasets, the host genetic association analyses were conducted in Chinese cohorts, so replication in other populations also needs to be assessed. Prediction of individual treatment responses and preventive effects against vaginitis or preterm birth were not evaluated in this study.
The study design and main findings are available in the Nature Genetics article.
Nature Genetics, Published online: 11 June 2026. DOI: 10.1038/s41588-026-02639-2
Summary: Bypassing the low prediction velocities and macro-environmental confounding errors that historically cloud empirical microbiome profiling in gynecology, this multi-centric study implements a programmable host-microbiome genomics infrastructure. Coupling high-resolution shotgun metagenomics with genome-wide association analyses (GWAS) across a multi-ethnic cohort of thousands of individuals, the computing platform establishes that distinct host genetic loci govern the spatial-temporal composition of the vaginal microbiome. The model deciphers the precise mathematical covariance linking continuous ancestral lineage variants to population-specific microbial abundance velocities, isolating loci associated with mucosal receptor affinity and innate immune signaling. This molecular calibration delivers a validated, non-invasive computational baseline to eliminate sample-overlap artifacts, optimize engineered live biotherapeutic product (LBP) docking profiles, and guide prospective universal patient stratification under digital metagenomic governance.
Researchers analyzing microbial sequences from vaginal samples can use GVMG as a reference to investigate microbes missed by existing databases and genomic differences within the same species. For example, when comparing microbial distributions across populations, follow-up analyses could examine strain-level differences in functional genes as well as species relative abundance. The replicated host genetic associations can also help prioritize functional experiments.
The paper states that the genomic data and summary statistics for the microbial genetic association analyses are publicly available. Applying them to select probiotics for individuals or predict disease risk would require separate performance assessments and clinical validation. For now, their practical value lies in providing a research reference resource and hypotheses to test.