Machine learning and Voronoi-based decision boundaries for Bacterial vaginosis to determine population- specific microbial interactions.

Support vector machine and random forest models accurately predict Bacterial vaginosis with balanced accuracy, identifying key Lactobacilli and anaerobic bacteria.

• Why it matters: Understanding how shifts in vaginal microbial populations lead to BV is crucial for improving diagnostics and treatment, yet the specific bacterial interactions remain unclear.
• What they did: Researchers compared multiple machine learning architectures and feature selection methods on 16s rRNA data from BV patients, then used explainable AI and Voronoi-based boundaries to analyze bacterial interactions.
• The result: They identified four Lactobacilli spp and six anaerobes as critical for BV diagnosis, providing new insights into microbial roles and offering a visual diagnostic framework based on bacterial abundance relationships.