Predicting Autism Spectrum Disorder Using Blood-based Gene Expression Signatures and Machine Learning.

Area of Science:

• Genomics
• Biomarkers
• Neurodevelopmental Disorders

Background:

• Autism Spectrum Disorder (ASD) diagnosis relies on behavioral observation, lacking objective biological markers.
• Gene expression profiling offers a potential avenue for identifying objective biomarkers for ASD.

Purpose of the Study:

• To identify a transcriptomic signature in peripheral blood for classifying individuals with ASD.
• To explore the potential of gene expression profiles as diagnostic biomarkers for ASD.

Main Methods:

• Utilized microarray data (GSE26415) from 21 young adults with ASD and 21 controls.
• Identified 19 differentially expressed probes using limma package in R.
• Validated probes using machine learning algorithms (SVM, k-NN) on a test dataset.

Main Results:

• Hierarchical clustering demonstrated good discrimination between ASD subjects and controls.
• Machine learning analysis achieved 93.8% overall prediction accuracy.
• Achieved 100% sensitivity and 87.5% specificity in classifying ASD cases.

Conclusions:

• Peripheral blood gene expression profiles can identify a biological signature for ASD.
• This exploratory study suggests potential for blood-based diagnostic biomarkers.
• Further validation with larger cohorts is necessary to enhance diagnostic accuracy.