Judith Van de Water, Ph.D.
University of California, Davis
NIEHS-funded researchers identified patterns in maternal autoantibodies, immune proteins that attack a person’s own proteins or tissues, that were highly associated with the diagnosis and severity of autism spectrum disorder (ASD). This is the first study to use machine learning to identify with 100% accuracy patterns specific to maternal autoantibody-related ASD (MAR ASD), which may serve as potential biomarkers of ASD risk. MAR ASD accounts for around 20% of all autism cases.
The team previously identified maternal autoantibodies against eight key proteins related to ASD. In the current study, the researchers sought to develop an accurate test for reactivity patterns against those proteins that could predict ASD. They analyzed plasma samples from 450 mothers of children with autism and 342 mothers of typically developing children from the Childhood Autism Risks from Genetics and Environment (CHARGE) study.
To train a machine learning algorithm, they identified autoantibody reactivity patterns associated with ASD. Next, they tested the accuracy of the patterns and built a high-precision model to predict ASD outcome. The machine learning program analyzed nearly 10,000 patterns and identified three top patterns associated with ASD.
Autoantibody reactivity to a single protein did not correlate with ASD diagnosis. Instead, reactivity to a combination of two or more proteins was needed for an association with ASD. For example, mothers with autoantibodies to CRIMP1 and GDA, two proteins involved in brain cell development, had 31 times greater odds of having a child with autism than the general population. In fact, autoantibodies to CRIMP1 combined with any of the top proteins increased the likelihood of a higher ASD severity score. According to the authors, this approach may aid early ASD detection, which can help facilitate early interventions.
Citation: Ramirez-Celis A, Becker M, Nuno M, Schauer J, Aghaeepour N, Van de Water J. 2021. Risk assessment analysis for maternal autoantibody-related autism (MAR-ASD): a subtype of autism. Mol Psychiatry; doi: 10.1038/s41380-020-00998-8. [Online 22 Jan 2021].