Early Risk Assessment of Autism Spectrum Disorder: A Novel Approach Using Microbial Biomarkers and Ensemble Classification Models

Overview

Autism Spectrum Disorder (ASD) is a prominent neurodevelopmental disorder that affects 1 in 36 children. This disorder is rapidly increasing in prevalence; however, it still lacks a reliable and objective diagnostic test. Clinical diagnostic criteria rely on subjective behavioral assessments, causing delays in intervention. This study hypothesizes that children with ASD have differentially abundant bacterial species in their gut microbiome compared to neurotypical controls using stool samples. If the gut microbiome composition differs between ASD and control subjects, it could suggest a potential gut -brain axis connection in ASD and identify potent ial biomarkers for diagnosis or therapeutic targets. Using 16S rRNA sequencing data from stool samples of ASD and neurotypical control subjects, a comprehensive analysis was performed to identify differentially abundant bacterial taxa. The workflow included: data preprocessing and quality control, taxonomic classification of OTUs, and statistical analysis using differential abundance testing (Mann Whitney Utest) to identify differences between ASD and control groups. Using the identified biomarkers, an ensemble machine learning classification model was developed to objectively assess the risk of ASD. The hypothesis was supported as the findings revealed significant differences between the ASD and control groups including an increased abundance of Prevotella and Lachnoclostridium species, and a decreased abundance of beneficial bacteria like Ruminococcaceae and Bacteroides, with p-values < 0.0005 indicating statistical significance. Furthermore, the machine learning model trained with these biomarkers had a prediction accuracy of 0.90 +/-0.05 and ROC of 0.95 +/ -0.02. This research can pave way for early diagnosis and targeted interventions for ASD. Connecticut

Competition history

• JSHS 2025 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.