“Towards the Development of a Diagnostic Test for Autism Spectrum Disorder: Data Science Meets Metabolomics”

Hosted by the Department of Biomedical Engineering

Abstract: 

Autism Spectrum Disorders (ASD) are a group of neurological disorders that present with limited social communication/interaction and restricted, repetitive behaviors/interests [1]. The current estimate is that approximately 2.8% of children in the US are diagnosed with ASD [2]. While this is a high prevalence and the economic burden by ASD is significant [3], there is still considerable debate regarding the underlying pathophysiology of ASD. Because of this lack of biological knowledge, autism diagnoses are restricted to observational behavioral and psychometric tools. Subsequently, the average age at which children receive an ASD diagnosis is four years [2], while it is generally acknowledged that diagnosis between 18-24 months is possible [4]. Furthermore, disparities by race/ethnicity in estimated ASD prevalence as well as disparities in the age of earliest comprehensive evaluation and presence of a previous ASD diagnosis or classification, suggest that access to treatment and services might be lacking or delayed for some children [2]. Thus, confirmation and expansion of the unique metabolic abnormalities in children with autism that accurately distinguishes them from typically-developing children would not only strengthen diagnostic accuracy, but also provide insights into underlying pathophysiology and a personalized approach to treatment options.

Stepping towards this goal of incorporating biochemical data into ASD diagnosis, we analyzed measurements of metabolite concentrations of the folate-dependent one-carbon metabolism and transulfuration pathways taken from blood samples of 83 participants with ASD and 76 age-matched typically-developing peers. Fisher Discriminant Analysis enabled multivariate classification of the participants as on the spectrum or typically-developing which results in 96.1% of all typically-developing participants being correctly identified as such while still correctly identifying 97.6% of the ASD cohort [5]. Furthermore, kernel partial least squares was used to predict adaptive behavior, as measured by the Vineland Adaptive Behavior Composite score, where measurement of five metabolites of the pathways was sufficient to predict the Vineland score with an R2 of 0.45 after cross-validation. This level of accuracy for classification as well as adaptive behavior prediction far exceeds any other approach in this field.

In addition to classification and adaptive behavior prediction, we have developed a steady state model of the metabolic pathways. We estimated flux parameters for every participant of the study and derived a probability density function for each parameter for the neurotypical participants and, separately, for participants with an ASD diagnosis. Comparison of the parameter distributions between neurotypical study participants and those on the autism spectrum revealed significant differences for four model parameters [6]. Sensitivity analysis identified the parameter describing the rate of glutamylcysteine synthesis, the rate-limiting step in glutathione production, to be particularly important in determining steady-state metabolite concentrations.

These computational studies enhance the analysis obtained from traditional population-level statistics and suggest that folate-dependent one carbon metabolism and transsulfuration may play an integral role in ASD pathophysiology. Although these results need to be replicated in independent studies, these analyses suggest combinations of metabolites in these pathways as potential biomarkers for ASD.

Registration is required to attend virtually. Please register.

Full schedule, BME Seminars

For inquries contact Mimi Hilburg.