Microbial dysbiosis and predictive neuroactive functional profiles in paediatric neurogenic bladder: an exploratory machine learning-based approach.

BACKGROUND: Neurogenic bladder (NB) in children is associated with impaired bladder emptying, recurrent urinary tract infections (rUTIs), and progressive upper urinary tract deterioration. Although urinary microbiome dysbiosis has been increasingly recognised in NB, its neuroactive functional potential and relationship to disease severity remain poorly understood. We investigated urinary microbial composition and predicted neuroactive metabolic pathways and evaluated their ability to stratify disease severity using integrative multivariate and machine learning (ML) analyses. METHODS: This cross-sectional study included 83 children comprising 39 patients with NB and 44 age- and sex-matched healthy controls. Patients with NB were stratified into Subgroup A (renal deterioration; n = 15) and Subgroup B (without renal deterioration; n = 23) based on glomerular filtration rate and renal scarring (DMSA). The urinary microbiome was characterised by 16S rRNA gene sequencing, and functional potential was inferred using PICRUSt2. A total of 25 literature-curated neuroactive MetaCyc pathways were analysed to generate a composite Neuroactive Dysbiosis Index (NDI). Associations between microbial taxa and predicted pathways were explored using correlation, hierarchical clustering, and network analyses. Partial least squares discriminant analysis (PLS-DA) and Random Forest (RF) models were used to identify discriminatory microbial signatures and evaluate classification performance. RESULTS: NB exhibited marked urinary microbial dysbiosis accompanied by predicted shifts in neuroactive metabolic pathways. Disease-associated microbiomes demonstrated predictive enrichment of pathways involved in amino acid catabolism, polyamine metabolism, coenzyme A and phospholipid biosynthesis, one-carbon metabolism, and central energy metabolism, together with depletion of pathways associated with microbial metabolic homeostasis and neurotransmitter precursor biosynthesis. The NDI demonstrated a progressive increase with disease severity (Control < Subgroup B < Subgroup A), with significant differences between Subgroup A and controls (p = 0.004) and between Subgroup A and Subgroup B (p = 0.002). Network analysis revealed extensive reorganisation of genus-pathway interactions in deteriorated NB, whilst PLS-DA demonstrated clear functional separation between study groups. RF models achieved strong discrimination using predicted neuroactive pathways alone [area under the curve (AUC) = 0.882], which further improved when combined with microbial taxonomic features (AUC = 0.981), identifying phospholipid biosynthesis, coenzyme A metabolism, aromatic amino acid metabolism, and tRNA charging as the most informative discriminatory pathways. CONCLUSIONS: Paediatric NB is characterised by coordinated urinary microbial dysbiosis and alterations in predicted neuroactive functional pathways that parallel disease severity. Integrating microbial taxonomic composition with predicted functional profiles substantially improved disease classification, highlighting the potential of computationally inferred neuroactive microbial signatures for disease stratification. These findings provide new insights into the microbiome-neuroactive axis in paediatric NB.