Machine Learning Recognizes Patterns of Maternal Autoantibodies Associated to Autism

New Technology Identified Biomarkers Linked with a Subtype of Autism with 100% Accuracy 

An exciting new study from UC Davis’ MIND Institute has identified several patterns of maternal autoantibodies that are distinctly associated with autism. Through an artificial intelligence generated process called machine learning, the study’s authors investigated maternal autoantibody-related autism spectrum disorder (MAR ASD), which is thought to make up approximately 20% of autism cases. The research involved obtaining plasma samples from mothers enrolled in the CHARGE study. Samples from 450 mothers of children with autism were compared to samples of 342 mothers of non-affected children. The comparison detected reactivity to eight different proteins that are abundant in the fetal brain. The MIND Institute researchers hypothesized that using a machine learning algorithm could determine which autoantibody patterns were specifically associated with autism. Their theory was correct. The authors report that machine learning was able to identify MAR-ASD specific patterns as potential biomarkers of autism risk with 100% accuracy. This research could lead to future tests that can identify MAR-autoantibodies making it possible for women to understand the risk of giving birth to a child who may later develop autism. 

Original Article

Original Study

Show Buttons
Hide Buttons