Background. Microbial genes that produce or transform neuroactive compounds may contribute to autism heterogeneity, but their relation to dimensional symptoms is unclear. Objective. To test whether neuroactive microbial pathways selected a priori were associated with symptom severity in autistic children and improved out-of-sample prediction beyond clinical covariates. Methods. This secondary cross-sectional analysis used linked metagenomic and clinical data from 818 children (464 autistic). Six of twelve a priori neuroactive MetaCyc pathways, detected in at least 10% of autistic participants, were analyzed. Social Responsiveness Scale, Second Edition (SRS-2), Anxiety Scale for Children–Autism Spectrum Disorder (ASC-ASD), and Child Behavior Checklist (CBCL) Internalizing scores were regressed robustly on centered log-ratio pathway abundance, age, sex, and batch, with Benjamini–Hochberg correction within outcome. Sensitivity analyses varied covariates and transformation, modeled pathways jointly, and examined influential observations, sparse detection, and cross-validated ridge prediction. Results. Sample sizes were 464 (SRS-2), 459 (ASC-ASD), and 144 (CBCL Internalizing). PWY-5088 was inversely associated with SRS-2 (β = −0.110, 95% CI −0.200 to −0.020; p = 0.017); the predefined false-discovery threshold was narrowly missed (q = 0.102), and the estimate persisted in sensitivity analyses. No pathway was consistently associated with ASC-ASD or CBCL Internalizing. Pathways added minimal SRS-2 prediction gain (mean ΔR² = 0.004) and worsened ASC-ASD prediction (mean ΔR² = −0.022). Conclusion. Neuroactive pathway profiles were not broad or clinically useful correlates of symptom severity. A small inverse association between PWY-5088 and social-communication severity merits replication with longitudinal metabolomics; it is neither a diagnostic biomarker nor evidence of causality.
Copyrights © 2026