Developing a Predictive Gene Classifier for Autism Spectrum Disorders Based upon Differential Gene Expression...

Area of Science:

• Genetics
• Neuroscience
• Biomarker Discovery

Background:

• Autism spectrum disorders (ASD) are diagnosed behaviorally, with limited genetic markers applicable to most cases.
• Current genetic research identifies abnormalities in up to 20% of ASD cases, but no single gene applies to more than 1-2%.

Purpose of the Study:

• To apply class prediction algorithms to gene expression profiles for identifying ASD subtypes.
• To distinguish between phenotypic subgroups of idiopathic autism and non-autistic controls using gene expression data.

Main Methods:

• Utilized gene expression profiles from lymphoblastoid cell lines (LCL) of ASD subgroups and controls.
• Employed cluster analyses of behavioral severity scores and support vector machine (SVM) algorithms with leave-one-out validation.
• Validated classifier genes using high-throughput quantitative nuclease protection assays on new LCL samples.

Main Results:

• Achieved up to 94% classification accuracy distinguishing ASD subgroups from controls using limited gene sets.
• Demonstrated high sensitivities and specificities (~90% or better) in initial SVM analyses.
• Achieved ~82% overall prediction accuracy with ~90% sensitivity and 75% specificity in validation assays.

Conclusions:

• Gene expression panels derived from phenotypically homogeneous ASD subgroups may serve as useful biomarkers for diagnosing autism subtypes.
• Further validation with larger cohorts and confirmation in primary cells earlier in development are necessary for clinical translation.