The vaginal microbiome plays a crucial role in women’s health, with Lactobacillus species being key to maintaining a stable and protective environment. Vaginal microbiota transplants (VMTs) from healthy donors are emerging as a promising therapeutic approach to restore microbial balance in women with vaginal dysbiosis. However, assessing engraftment and treatment efficacy remains a challenge due to the complex mixtures of strains and species involved in VMTs.
Traditional approaches such as phylogenetic tree-based analyses are limited to detecting dominant strains and often fail to resolve strain-level dynamics in mixed communities. To address this, we developed CHAMP™ StrainQ, a machine learning-based method that leverages single nucleotide variant (SNV) profiles to accurately detect and track multiple strains of the same species within a sample.
We demonstrate that CHAMP™ StrainQ can reliably resolve mixtures of up to seven probiotic Lactobacillus crispatus strains in the presence of up to seven endogenous strains, achieving a 131% improvement in sensitivity compared to existing methods. The tool was validated using in silico-generated strain mixtures and has been successfully applied in published studies involving vertical strain transmission and probiotic engraftment in infants and adults.