The main features of ASD are stereotyped behavior and impaired communication, but may exhibit many other symptoms such as hypersensitivity to sound and integrated motor injuries. Although it has been suggested that genetic factors contribute strongly to the increased risk of ASD, more than 1,000 candidate genes have been reported so far, and genetic susceptibility factors have not been identified.

　Under these circumstances, in this study, we use machine learning, which is an artificial intelligence technology, to perform cluster analysis from various symptoms exhibited by ASD and combine it with genome-wide association study (GWAS) to identify genetic susceptibility factors. I hypothesized that the number would increase.

　In fact, while no significant association was observed with the GWAS method using the "all patient groups" and the target group, which were grouped under the disease name ASD, ASD patients were grouped (clustered) and "per cluster". In GWAS conducted in the "patient group" and the target group, 65 significant loci were identified.In other words, ASD may be a collection of heterogeneous diseases, and by clustering cases into a more homogeneous population, personalized medicine can be expected to be realized according to the characteristics of each population.

　This result is related to elucidating the genetic structure and pathogenesis of ASD, providing clues to promote the development of precision medicine for ASD, and analyzing disease groups in clusters using machine learning techniques. The finding that it is easy to find genes is expected to contribute to the progress of personalized medicine for many diseases in the future.

Paper information:[Translational Psychiatry] Clustering by phenotype and genome-wide association study in autism