Development of a Predictive Gene Classifier for Autism Spectrum Disorders Based Upon Differential Gene Expression Profiles Between Cases and Controls

Background:  Autism is a neurodevelopmental disorder which is currently diagnosed solely on the basis of abnormal behavior as well as observable deficits in communication and social functioning.  Although a variety of autism candidate genes have been identified on the basis of genetic analyses, none have been shown to be unequivocally diagnostic for idiopathic autism or to account for more than a few % of autism cases.

Objectives:   To identify limited sets of differentially expressed genes which can robustly distinguish between autistic cases and controls as potential biomarkers for diagnostic screening

Methods:   DNA microarray analysis was employed to obtain the gene expression profiles of lymphoblastoid cell lines (LCL) of 87 autistic male individuals who were divided into 3 phenotypic subgroups (Hu et al., Autism Research, 2:78-97, 2009) based on cluster analyses of severity scores on the Autism Diagnostic Interview-Revised (ADI-R) assessment instrument (Hu and Steinberg, Autism Research, 2:67-77, 2009).  We compared these expression profiles against that obtained from LCL of 29 nonautistic male control subjects.  Using these datasets, we then utilized gene classification and cross-validation analysis software to identify sets of differentially expressed genes that have a high statistical probability of predicting cases and controls. 

Results:   We have identified panels of selected genes (less than 30) which correctly classify samples according to affected/unaffected status with an accuracy exceeding 90%.  When autistic samples are subtyped according to ADI-R cluster analyses prior to the gene expression and classification analyses, the accuracy of correct assignment to cases and controls exceeds 98%.  High throughput quantitative nuclease protection assay of a subset of “classifier” genes (n=14) for one of the ASD subtypes further confirms the ability of the selected differentially expressed genes to identify autistic and control subjects with an accuracy of ~80%.

Conclusions:   Limited sets of differentially expressed genes can distinguish between autistic cases and controls with high accuracy.  We suggest that such panels of genes may serve as useful biomarkers for screening or diagnosis of idiopathic autism.