Harnessing vaginal inflammation and microbiome: a machine learning model for predicting IVF success

Humans are the only species with a commensal Lactobacillus-dominant vaginal microbiota. Reproductive tract microbes have been linked to fertility outcomes, as has intrauterine inflammation, suggesting immune response may mediate adverse outcomes. In this pilot study, we compared vaginal microbiota composition and immune marker concentrations between patients with unexplained or male factor infertility (MFI), as a control. We applied a supervised machine learning algorithm that integrated microbiome and inflammation data to predict pregnancy outcomes. Twenty-eight participants provided vaginal swabs at three IVF cycle time points; 18 achieved pregnancy. Pregnant participants had lower microbial diversity and inflammation. Among them, MFI cases had higher diversity but lower inflammation than those with unexplained infertility. Our model showed the highest prediction accuracy at time point 2 of the IVF cycle. These findings suggest that vaginal microbiota and inflammation jointly impact fertility and can inform predictive tools in reproductive medicine.