💡 Autism Spectrum Disorder (ASD) is characterized by neurodevelopmental deficits, with emerging evidence implicating the gut microbiome in its pathophysiology. This study employs a machine learning approach to identify bacterial taxa associated with ASD and validate their predictive power across multiple datasets.
📍 Using 16S rRNA gene sequencing data from ASD cases and siblings, 26 bacterial taxa are identified as discriminators of ASD status, with an average area under the curve (AUC) of 81.6%. Validation in independent cohorts yields AUCs of 74.8% and 74%, indicating robustness. These findings underscore the potential of the gut microbiome as a diagnostic and therapeutic target for ASD.
📍 Key Findings:
📌 Machine Learning Analysis: Recursive Ensemble Feature Selection (REFS) applied to 16S rRNA gene sequencing data identifies 26 bacterial taxa distinguishing ASD cases from controls with an average AUC of 81.6% in a sibling-controlled dataset.
📌 Validation Across Cohorts: The selected bacterial taxa demonstrate predictive power in two independent cohorts, with average AUCs of 74.8% and 74%, supporting the robustness of the findings.
📌 Multivariate, Multidataset Approach: Integration of feature selection and cross-validation strategies overcomes limitations of traditional statistical methods, yielding highly accurate predictive models for ASD classification.
📌 Bacterial Taxa Associations: Differential abundances of bacterial taxa, including decreased 𝘉𝘪𝘧𝘪𝘥𝘰𝘣𝘢𝘤𝘵𝘦𝘳𝘪𝘶𝘮 and increased 𝘊𝘭𝘰𝘴𝘵𝘳𝘪𝘥𝘪𝘢, 𝘚𝘢𝘳𝘤𝘪𝘯𝘢, 𝘢𝘯𝘥 𝘗𝘢𝘳𝘢𝘣𝘢𝘤𝘵𝘦𝘳𝘰𝘪𝘥𝘦𝘴, are observed in ASD cases compared to controls. These findings align with previous studies and suggest potential metabolic pathways implicated in ASD pathophysiology.
📌 Biological Significance: Despite limitations such as variable ASV lengths and confounding factors, the identified bacterial taxa offer insights into gut-brain axis interactions in ASD. Targeting the gut microbiome with dietary interventions may hold promise for improving ASD-related symptoms.
📍 This study identifies a robust gut microbiome signature for ASD classification across multiple cohorts, highlighting the potential of machine learning approaches in microbiome research. The identified bacterial taxa provide valuable insights into ASD pathophysiology and offer avenues for diagnostic and therapeutic development.
Link to the article : http://tinyurl.com/3mn7ean2
