Comparative analysis of gut microbiota structure in the Xinjiang (China) population of children with mild-to-moderate and severe autism spectrum disorder

Background Autism spectrum disorder (ASD) exhibits substantial clinical heterogeneity; however, the relationship between gut microbiome features and ASD severity remains incompletely understood. Methods We enrolled 88 children with ASD from Xinjiang, China, and characterized gut microbial profiles using 16S rRNA gene sequencing. α and β diversity were compared between children with mild-to-moderate ASD (MM-ASD) and severe ASD (S-ASD), and their associations with Childhood Autism Rating Scale (CARS) scores were assessed. Differential abundance analyses were performed using ANCOM-BC2, LinDA, and LEfSe. Predicted functional potentials were inferred using PICRUSt2. A nested cross-validation framework with random forest models was applied to evaluate the discriminative performance of microbial features. Results No significant differences were observed in overall α or β diversity between the MM-ASD and S-ASD groups, and CARS scores were not significantly associated with microbial diversity metrics. By integrating results from ANCOM-BC2, LinDA, and LEfSe, five candidate genera with consistent enrichment patterns were identified: Angelakisella and CAG-83 were enriched in the S-ASD group, whereas Terrisporobacter, Lactobacillus, and Adlercreutzia were enriched in the MM-ASD group. The relative abundances of Angelakisella and CAG-83 showed positive correlations with CARS scores, while Adlercreutzia exhibited a negative correlation with CARS scores. PICRUSt2-based analysis suggested differences in predicted energy metabolism-related functional potentials; however, these functional inferences remain hypothetical and require validation using direct functional assays. In nested cross-validation, the microbial feature-based model demonstrated limited discriminative performance (AUC = 0.579) compared with the baseline model (AUC = 0.521). Conclusion In this small, single-region, cross-sectional cohort, gut microbial characteristics were not robustly associated with overall ASD severity. Future multicenter, larger-scale longitudinal studies integrating multi-omics approaches and standardized clinical assessments are warranted to further clarify the relationship between gut microbiome features and ASD severity.