Back to 2026 Abstracts
Urinary Microbiome Dysbiosis in Children with Neurogenic Bladder Associated with Upper Tract Deterioration- A Comprehensive Microbiome Analysis including Supervised Machine Learning-Based Approach
Sachit Anand, MBBS, MCh, Deephi R. Ramya, M.Sc,, Ajay Verma, MBBS, MCh, Jitendra Kumar Meena, MBBS, MD, Himalaya Kumar, M.Sc,, Pankaj Hari, MBBS, MD, FIAP, FAMS, Vineet Ahuja, MBBS, MD, DM, MNAMS.
All India Institute of Medical Sciences, New Delhi, India.
BACKGROUND: Neurogenic bladder (NB) in children is a common cause of recurrent urinary tract infections, progressive renal scarring, and deterioration of kidney function. Standard culture-based diagnostics provide limited insight into chronic microbial disturbances or their relationship with disease severity. The urinary microbiome represents a potential mechanistic link between bladder dysfunction and renal outcomes, yet comprehensive characterization in pediatric NB populations, particularly regarding severity stratification, remains lacking. This study aims to characterize urinary microbiome dysbiosis in pediatric NB and evaluate its relationship with disease severity (upper tract deterioration), and train supervised machine-learning models for microbiome-based disease severity stratification.
METHODS: This single-center cross-sectional study enrolled 83 children, 39 with NB and 44 age- and sex-matched healthy controls. NB patients were stratified into subgroup A (renal function deterioration; GFR <60 mL/min/1.73m² or DMSA scans-confirmed scarring; n=15) and subgroup B (no deterioration; n=23). V3-V4 16S rRNA gene sequencing was performed using Illumina NovaSeq. Alpha diversity was assessed using observed ASVs, Chao1, Shannon, and Simpson indices; beta diversity was evaluated using Bray-Curtis and Jaccard distances with PERMANOVA. Differential abundance analysis employed DESeq2 with Benjamini-Hochberg correction. Functional potential was inferred using PICRUSt2 mapped to MetaCyc pathways, and antimicrobial resistance profiles were characterized using CARD/RGI aligned to WHO AWaRe categories. Three supervised machine-learning (ML) models (Random Forest, XGBoost, LightGBM) were independently trained using prevalence-filtered taxa with one-vs-rest classification, evaluated by ROC-AUC with cross-validation.
RESULTS: Children with NB demonstrated significantly reduced microbial diversity compared with controls: Chao1 richness (W=1513, p=1.0×10⁻¹⁰), observed ASVs (W=1529, p=9.5×10⁻¹⁰), and Shannon diversity (W=1288, p=5.9×10⁻⁵). A severity-dependent gradient emerged (subgroup A < subgroup B < control). Beta diversity analyses confirmed significant compositional separation (Bray-Curtis PERMANOVA R²=0.021, p=0.001), with greatest dispersion in subgroup A, consistent with the Anna Karenina principle of microbiome dysbiosis. Taxonomically, NB was characterized by enrichment of pathobionts (Escherichia-Shigella, Enterococcus, Corynebacterium, Pseudomonas, Acinetobacter) and depletion of commensal genera (Porphyromonas, Dialister, Prevotella). Functional inference revealed severity-linked enrichment of glycolytic, anaerobic respiratory, methylation, and redox-balancing pathways, with concurrent depletion of oxidative and biosynthetic functions. Resistome analysis demonstrated expanded antimicrobial resistance profiles in NB, particularly to WHO Reserve-category antibiotics. Among ML models, Random Forest and XGBoost achieved excellent discriminatory performance (AUC=0.99 for control classification), with SHAP-based severity scores demonstrating monotonic gradients across clinical groups. Decision curve analysis confirmed the superior net clinical benefit of microbiome-guided strategies over treat-all approaches.
CONCLUSIONS: Pediatric NB is characterized by severity-dependent urinary microbiome dysbiosis marked by progressive diversity loss, increased inter-individual variability, pathobiont enrichment, and functional reprogramming toward stress-adaptive metabolism. The trained ML-based models showed excellent discriminatory performance for genus-based disease severity stratification, supporting potential clinical utility for non-invasive risk assessment and personalized management in pediatric NB in future studies.
Back to 2026 Abstracts
