Unraveling the Dysbiosis of Vaginal Microbiome to Understand Cervical Cancer Disease Etiology-An Explainable AI ApproachOriginal paper
Karen Pendergrass is a microbiome researcher specializing in microbiome-targeted interventions (MBTIs). She systematically analyzes scientific literature to identify microbial patterns, develop hypotheses, and validate interventions. As the founder of the Microbiome Signatures Database, she bridges microbiome research with clinical practice. In 2012, based on her own investigative research, she became the first documented case of FMT for Celiac Disease, four years before the first published case study.
Read MoreWhat was studied?
Researchers analyzed the vaginal microbiome to understand its role in cervical cancer development. They compared microbial composition between cervical cancer samples and healthy controls.
How was it studied?
Relative abundance, diversity, and LEfSe analyses characterized the bacterial communities. A random forest model with repeated k-fold cross-validation was trained on the data, then SHapley Additive exPlanations (SHAP) were used to identify which taxa most influenced the model's predictions.
What did they find?
Firmicutes, Actinobacteria, and Proteobacteria dominated at the phylum level, with Lactobacillus iners and Prevotella timonensis significantly increased at the species level in cancer samples. Cervical cancer samples showed reduced diversity, richness, and dominance versus controls. LEfSe linked Lactobacillus iners along with Lactobacillus, Pseudomonas, and Enterococcus genera to cervical cancer, and functional enrichment tied the community shifts to aerobic vaginitis, bacterial vaginosis, and chlamydia. SHAP identified increased Ralstonia as the strongest predictor of cervical cancer status in the model.
Why it matters
The explainable AI approach uncovered Ralstonia as a new candidate microbial marker in cervical cancer, beyond the more studied Lactobacillus shifts, suggesting fresh targets for dysbiosis-linked cancer research.