Maternal prenatal gut microbiome and domain-specific neurodevelopment in early childhood: associations and exploratory prediction

Wei, Hongcheng, Jin, Bowen, Xu, Yifan, Zhao, Xiaoyu, Ning, Jie, Gao, Beibei, Zhang, Mingzhi, Tun, Hein Min, Xia, Yankai
Science China Life Sciences · Springer Nature · 2026-11-17 · DOI: 10.1007/s11427-025-3551-7

Аннотация

The potential of the maternal prenatal gut microbiome for prediction of domain-specific neurodevelopment in children has not been fully appreciated. We sought to explore the associations and predictive capacity of the maternal prenatal gut microbiome in relation to children’s domain-specific neurodevelopment. Leveraging a prospective birth cohort, we performed both 16S rRNA gene amplicon sequencing and shotgun metagenomic sequencing for pregnant women’s fecal samples. The domain-specific neurodevelopment in children was evaluated utilizing the Ages and Stages Questionnaire (ASQ). We used the prenatal gut microbiome at the phylum, genus, species, and pathway levels in combination with multiple machine learning models, including random forest, extreme gradient boosting (XGBoost), light gradient-boosting machine (LightGBM), least absolute shrinkage and selection operator (LASSO) logistic regression, and ridge logistic regression, individually and in conjunction with conventional risk factors to predict children’s neurodevelopment. We found that the gut microbiome features, individually or integrated with conventional risk factors, improved the predictive performance of machine learning models for children’s neurodevelopment across five subdomains, achieving area under the receiver operating characteristic curves (AUROCs) up to 0.88. Compared with AUROCs of 0.70–0.76 for communication, 0.70–0.72 for gross motor, 0.71–0.77 for fine motor, 0.68–0.77 for problem solving, and 0.57–0.62 for personal-social development using conventional risk factors alone, the combined models achieved AUROCs of 0.79–0.83, 0.71–0.80, 0.82–0.84, 0.79–0.88, and 0.59–0.64, respectively. The highest area under the precision-recall curves (AUPRCs) were 0.63, 0.50, 0.63, 0.60, and 0.39 for communication, gross motor, fine motor, problem solving, and personal-social development, respectively, compared with prevalence-matched AUPRC baselines of 0.12, 0.14, 0.12, 0.12, and 0.17. However, performance gains varied across algorithms and feature configurations. Maternal prenatal gut microbial features were associated with domain-specific neurodevelopment and showed exploratory predictive signals in selected internally validated models. Independent external cohorts and age-specific longitudinal validation are required before these features can be considered clinically useful biomarkers or used to inform early-identification or intervention strategies.

Читать у издателя →

Похожие статьи