Predicting autism spectrum disorder through the gut microbiota composition using machine learning

Abstract

The human gut microbiota plays a crucial role in overall health, impacting various diseases, including autism spectrum disorder (ASD). This study explores the relationship between gut microbiota changes and ASD by analyzing microbial compositions and abundances of 692 gut microbiota samples using public 16S rRNA sequencing datasets. Data preprocessing included normalization and redundancy reduction, retaining 367 microbial features. A machine learning model was then developed to predict ASD, utilizing feature-selected random forest algorithms that showed superior performance in both training and independent test sets. Identified microbial features with high correlation to ASD included Clostridiales bacterium VE202-08, Solobacterium moorei gene, and other features. The findings suggest that modulating the gut microbiota composition could mitigate ASD risk or alleviate symptoms. These insights pave the way for novel ASD diagnostic methods through microbiota analysis, although further research is required to validate these possibilities. This study offers a new perspective on the etiology and progression of ASD and proposes potential predictive tools for its diagnosis.