Machine Learning Discovers Potential Markers of Autism, which may reveal the pathogenesis.
According to a study published in Molecular Psychiatry, machine learning tools have identified the mechanism by which mothers produce antibodies. These patterns will reveal the likelihood and severity of children suffering from autism.
The research team pointed out that the incidence of autism spectrum disorder (ASD) has been rising, but there is still a lack of ASD risk biomarkers. According to the researchers, the Centers for Disease Control and Prevention (CDC) estimated that 59% of children in the United States had autism in 2018, which makes autism the most important health problem and affects affected families and medical systems. Brings a huge economic burden.
Autoantibodies are immune proteins that attack human tissues. Previously, the research team found that the pregnant mother’s autoantibodies can respond to the growing brain of the fetus after it has developed.
In the current study, the researchers analyzed the plasma samples of 450 mothers of children with autism and 342 typically developing children to test the responsiveness to eight different proteins abundant in the fetal brain. The research team then used machine learning algorithms to determine which autoantibody patterns are particularly relevant for the diagnosis of ASD.
The team developed and validated a test to identify patterns of ASD-specific maternal autoantibodies that are reactive against eight proteins that are highly expressed in the developing brain. This maternal blood test uses an enzyme-linked immunosorbent assay (ELISA) platform, which is fast and accurate.
The lead author of this study, Judy van der Water, professor of rheumatology, allergy, and clinical immunology at the University of California, Davis, said: “The significance of this particular study is that we provide for future clinical The app developed a new, translatable test.”
The machine learning algorithm analyzes about 10,000 patterns and the top of the logo is related to maternal autoantibodies related to Autism Spectrum Disorder (MAR ASD), that is, three patterns related to conditions that account for approximately 20% of all autism cases: CRMP1 + GDA, CRMP1 + CRMP2 and NSE + STIP1.
“For example, if a mother has autoantibodies to CRIMP1 and GDA (the most common pattern), then based on the current data set, she is 31 times more likely to have a child with autism than the general population. Huge,” Van der Water said. “There is almost nothing that can provide you with this type of risk assessment.”
The research team also found that in any of the highest modes, the response to CRMP1 significantly increased the odds of children with severe autism.
With these maternal biomarkers, it is possible to make early diagnosis of MAR autism and carry out more effective behavioral interventions.
“The significance of this research is huge,” Van der Water said. “This is the first time that machine learning has been used to identify specific patterns of MAR ASD with 100% accuracy as potential biomarkers of ASD risk.”
The study paves the way for more research on potential pre-pregnancy testing, which may be a viable option for high-risk women over the age of 35 or who have had children with autism.
“We can imagine that a woman can receive blood tests for these antibodies before pregnancy. If she has them, she will know that she is at high risk of having a child with autism. Otherwise, because MAR autism is ruled out, She has a 43% lower chance of having a child with autism.”
The team is currently using animal models to study the pathological effects of maternal autoantibodies.
Van de Water said: “We will also use these animal models to develop therapeutic strategies to prevent the fetus’s maternal autoantibodies.” “In terms of early risk assessment of autism, this research is of great significance, and we hope this technology It will be clinically useful in the future and Machine Learning Discovers Potential Markers of Autism.