Nat Med: new AI skills to help screen for autism

in China, the incidence rate of autism in children aged 6-12 is 0.7%. It is estimated that one in 54 children in the United States is autistic. At present, the diagnosis of autism is only based on symptoms, and should be differentiated from Asperger’s syndrome and generalized developmental disorder to be classified. But it is difficult to distinguish them according to their clinical symptoms alone. Recent genomic studies have shown that precision medicine in the United States has proposed that “when multiple molecular indicators are combined with conventional clinical, histological and laboratory results, they can help us to describe and classify diseases more accurately”. Precision medicine can make full use of the diversity of data. In addition to the absolute size of biomedical data, it can layer a variety of data patterns, and can identify subgroups of patients with pathophysiological similarity. < / P > < p > first, the researchers recruited 42 patients with ASD, collected 524 samples from 26 brain regions, and collected the wes data set of 3531 individuals from 1704 families through the national autism research database of the National Institute of mental health data archive. < p > < p > subsequently, large data sets of complete exome sequences, neurodevelopmental expression patterns, electronic health records, and healthcare statements were integrated. Furthermore, harmful mutations that regulate neurodevelopment, gender differential expression and ASD segregation were examined to identify genetic and possible gene damaging variants, namely nonsense mutation, frameshift mutation and splice site mutation. Multiple hypothesis tests were strictly controlled. Thus, 33 related exons were determined. Subsequently, the functional enrichment analysis of these exons revealed an unreported molecular polymerization, namely lipid regulation. For example, the study of small clusters of five exons of the LDL receptor gene carries harmful mutations isolated from ASD. These mutations are predicted to alter LDL, cholesterol and triglyceride levels. < / P > < p > the researchers also examined the prevalence of dyslipidemia in 80714 diagnosed ASD patients. The results showed that dyslipidemia was significantly increased in ASD patients, including total cholesterol, HDL, LDL and triglycerides. < / P > < p > finally, the researchers verified the relationship between dyslipidemia and ASD in mice. In short, by mining the mouse genomic informatics database, according to Mendel’s human genetic compendium, the phenotypes of mice with ASD and dyslipidemia targeted mutations were clustered. The results showed that the dyslipidemia mice model had social and nervous system abnormalities, and ASD mice also had dyslipidemia and abnormal growth. < / P > < p > in general, researchers combined the whole exon sequencing data of autistic children and their siblings, the family pedigree data of autistic children, the electronic medical record data of millions of people and the health insurance data of tens of millions of people, and animal model data to determine the subtypes of autism characterized by dyslipidemia. This study is of great significance for early screening and intervention of autism, suggesting that our routine blood lipid test can help to screen the risk of neonatal autism, and provide a new direction for early intervention of autism. Focus