A Manifold-Based Framework for Studying the Dynamics of the Vaginal Microbiome

Area of Science:

• Microbiome research
• Infectious disease dynamics
• Computational biology

Background:

• The vaginal microbiome's composition, categorized into community state types (CSTs), is vital for reproductive health.
• Lactobacillus-dominant CSTs are considered healthy, while non-Lactobacillus CSTs are associated with Bacterial vaginosis (BV), a condition increasing susceptibility to infections.
• The precise pathways and dynamics of the vaginal microbiome transitioning from a healthy to a BV state remain incompletely understood.

Approach:

• A manifold-detection framework, inspired by single-cell analysis, was applied to 8,026 vaginal samples.
• This approach identifies low-dimensional trajectories within the high-dimensional microbiome composition space.
• Samples were ordered along these trajectories to assign a pseudo-time score, indicating proximity to the BV state.

Key Points:

• Distinct progression routes from healthy CSTs to BV were identified, with pseudo-time scores correlating with microbial diversity and health status.
• BV indicators, such as Nugent score and Amsel's criteria, were accurately predicted using pseudo-time scores.
• Gardnerella vaginalis was identified as a key taxon in BV progression, with its abundance increasing along the identified unhealthy trajectories.

Conclusions:

• Manifold detection effectively characterizes the progression from healthy vaginal microbiomes to BV.
• This computational framework provides a quantitative measure of health status for new samples along the BV development continuum.
• The findings offer novel insights into the dynamics of vaginal dysbiosis and BV pathogenesis.