Researchers used machine learning on clinical data from 6,355 autistic children in the SPARK cohort, then validated the results in 2,717 more children. They identified three subtypes: high support needs, emotional-behavioral support needs, and low support needs. Each showed different patterns of genetics, medication use, and services.
Most autism research treats autism as one category, which hides meaningful differences. This framework links subtypes to specific genetic influences — common variants tied to ADHD and depression risk in one group, rare variant burden in another. That could eventually help families and clinicians anticipate which supports a child may need.
✦ Autism is not one thing; clearer subtypes could lead to better-matched support.