Human imaging data came from ABIDE and the Child Mind Institute, with findings replicated across dozens of independent sites to strengthen validity.

This cross-species, biologically anchored approach aims to lay groundwork for precision medicine and more targeted interventions in autism.

The study analyzed data from about 940 autistic individuals and over 1,000 neurotypical controls, with mouse models used to map connectivity to cellular and molecular processes.

Researchers caution that these subtypes likely represent only part of autism's diversity and expect more subtypes to emerge with larger datasets and methods.

Published in Nature Neuroscience, the work supports a biologically grounded stratification of autism through multidimensional analysis.

Findings were replicated across human data (roughly 940 with autism and 1,036 controls) and 20 autism-like mouse models, providing cross-species validation.

Researchers identify two brain-based subtypes of autism—hypoconnectivity and hyperconnectivity—based on distinct brain connectivity patterns linked to different genetic and immune-related pathways.

Approximately one quarter of analyzed autism brains fall into either subtype, suggesting these subtypes capture meaningful portions of autism biology but not all cases.

Hypoconnectivity shows reduced brain connectivity and is associated with synapse-related genes, while hyperconnectivity shows increased connectivity and links to immune-system genes, with the latter showing slightly more severe autism measures.

The dataset and analysis tools are openly available to the scientific community to foster replication and further investigation.

If clinically confirmed, subtyping could enable more targeted therapies and support programs beyond a one-size-fits-all approach.

Mouse models provide a biological framework translating to human scans, acting as a Rosetta Stone to connect molecular pathways with connectivity signatures.