The researchers used machine learning to find the maternal autoantibody patterns associated with autism
Using machine learning, researchers at UC Davis MIND have identified several types of maternal autoantibodies closely related to the diagnosis and severity of autism.
Their study was published on January 22 in Molecular Psychiatry, It places special emphasis on autism spectrum disorder associated with maternal autoantibodies (MAR ASD), a condition that accounts for about 20% of all cases of autism.
“The implications of this study are enormous,” said Judy Van de Water, professor of rheumatology, allergies and clinical immunology at the University of California, Davis and lead author of the study. “It is the first time that machine learning has been used to identify patterns of MAR ASD with 100% accuracy as potential biomarkers of ASD risk.”
Autoantibodies are immune proteins that attack a person’s own tissues. Previously, Van de Water found that a pregnant mother’s autoantibodies could interact with her developing fetus’ brain and alter its development.
Machine learning identifies patterns that indicate the likelihood and severity of autism
The research team obtained plasma samples from mothers enrolled in the CHARGE study. They analyzed samples from 450 mothers of autistic children and 342 mothers of developing children, also from CHARGE, to discover interactions with eight different proteins abundant in the fetal brain. Then they used a machine learning algorithm to identify the patterns of autoantibodies that were specifically associated with the diagnosis of ASD.
Researchers have created and validated a test to determine the reaction patterns of maternal autistic antibodies to autism against eight proteins that are highly expressed in the developing brain.
“What is important about this particular study is that we have created a new test that is translatable for future clinical use,” Van de Water said. This simple maternal blood test uses the ELISA platform (Enzyme-Linked Pipette Assay), which is very fast and accurate.
The machine learning software broke nearly 10,000 patterns and identified three main patterns associated with MAR ASD: CRMP1 + GDA, CRMP1 + CRMP2, and NSE + STIP1.
“For example, if a mother has autoantibodies to CRIMP1 and GDA (the most common pattern), then the chances of having a child with autism are 31 times more likely than the general population, based on the current data set.” Van de Water said. This gives you this kind of risk assessment. ”
The researchers also found that interacting with CRMP1 in any of the overhead types significantly increases a child’s odds of developing severe autism.
Van de Water notes that with these maternal biomarkers there are possibilities for a very early diagnosis of MAR autism and more effective behavioral intervention. The study opens the door for further research on possible pre-pregnancy testing, and it is especially useful for women at risk over the age of 35 or who have already given birth to a child with autism.
“We can imagine that a woman could have a blood test for these antibodies before pregnancy. If she had, she would know that she would be at a very high risk of having an autistic child. If not, she had a 43% less chance of having an autistic child. With autism where autism MAR is excluded.
Van de Water is currently investigating the pathogenic effects of maternal autoantibodies using animal models. “We will also use these animal models to develop treatment strategies to block the mother’s autoantibodies from the fetus,” Van de Water said.
“This study is very important with regard to early assessment of autism risk, and we hope this technology will become something clinically beneficial in the future.”
The study’s first author is Alexandra Ramirez Silis, with co-authors Joseph Schauer and Miriam Nino of the University of California, Davis, and Nima Agaibor and Martin Baker of Stanford University.
Funding for the study was provided by the NIEHS Center for Children’s Environmental Health and Environmental Protection Agency (EPA) (2P01ES011269-11, 83543201), NIEHS-funded CHARGE study (R01ES015359), NICHD-funded IDDRC 054 (U54HD079125), and the National Council for Science and Technology (CONACYT- UC MEXUS) PhD Fellowships, NIH Scholarships R35 GM138353.
http: // dx.