A study published in Nature Neuroscience has split autism into two biologically distinct subtypes based on how brain regions communicate with each other. Researchers mapped connectivity patterns across 20 autism-relevant mouse models and then matched those signatures to brain scans from 940 autistic and more than 1,000 neurotypical people. One subtype features abnormally weak connections between brain areas, while the other shows abnormally strong ones, and each aligns with a separate set of molecular pathways.
Why two wiring patterns change the clinical picture
Autism research and clinical practice have long treated the condition as a single diagnostic category, even though the range of traits, severity levels, and treatment responses varies enormously from person to person. That heterogeneity has frustrated drug developers and clinicians alike: interventions that help some autistic individuals show little effect in others, and no biological marker has reliably predicted who will benefit from what. Splitting the population into a hypoconnectivity group and a hyperconnectivity group, each tied to distinct gene-expression profiles, offers a concrete biological framework that behavioral assessments alone have never provided.
The practical stakes are immediate. If the two subtypes respond differently to therapies, matching the right approach to the right wiring pattern could shorten years of trial and error for families. One testable prediction: autistic individuals whose brains fall into the hyperconnectivity subtype may show faster gains in social cognition after targeted cognitive-behavioral training than those in the hypoconnectivity subtype, regardless of baseline IQ. That hypothesis has not been tested directly, but the biological split now gives researchers a measurable variable to design such trials around. Without that variable, clinical studies have been averaging results across what may be two fundamentally different conditions.
Cross-species brain mapping and the ABIDE dataset
The research team built its framework by first analyzing resting-state functional MRI data from 20 mouse models linked to autism-associated genes. Each model carried a different genetic mutation known to affect brain development, and the mice fell into two clusters: one with reduced connectivity between brain regions and another with heightened connectivity. The researchers then translated those mouse-derived signatures into human brain coordinates and applied them to a large, multi-site collection of human fMRI scans.
That human dataset came from the Autism Brain Imaging Data Exchange, a two-phase open-science effort. The first phase, known as ABIDE I, was released in 2012 and aggregated resting-state fMRI and phenotypic data from multiple institutions under shared quality-control standards. A second release, ABIDE II, expanded the collection with additional psychiatric variables and diffusion imaging datasets. Together, the two phases gave the team access to scans from 940 autistic participants and more than 1,000 neurotypical controls, drawn from sites around the world.
The two human subtypes that emerged mirrored the mouse clusters. One group showed broadly reduced functional connectivity, while the other showed broadly increased connectivity. Each subtype mapped onto different molecular pathways: one linked to synaptic signaling and the other to immune-related processes. The consistency across species and across scanning sites is what gives the finding its weight. Random noise or site-specific scanner artifacts would not produce the same two-cluster split in both mice and humans.
Gaps in subtype data and what to watch next
Several open questions limit how quickly these subtypes can change clinical decisions. The ABIDE phenotypic files do not include detailed medication histories or long-term symptom tracking, so researchers cannot yet confirm whether an individual’s subtype stays stable over months or years, or whether medications shift someone from one connectivity pattern toward the other. The published study also does not break down clinical severity scores by subtype in enough detail for outside teams to immediately design stratified drug trials.
A separate commentary in Nature Neuroscience noted that prior attempts to subtype autism through fMRI have struggled with replication, partly because head motion during scanning can mimic or mask real connectivity differences. The new study’s cross-species design sidesteps some of that concern, since mouse scans are acquired under anesthesia with minimal motion, but the human arm of the analysis still depends on standard motion-correction methods whose limits are well known.
The most concrete next step is a prospective clinical trial that recruits participants, assigns them to a subtype based on their fMRI profile, and then tracks whether the two groups respond differently to the same intervention. Until that trial runs, the two-subtype framework remains a powerful classification tool rather than a proven clinical guide. Families and clinicians should watch for announcements from research groups that gained early access to the subtype-assignment methods, because the first stratified pilot studies will determine whether this biological split translates into meaningfully different treatment pathways.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
