fMRI study defines two autism subtypes

Nature Neuroscience published a study on May 15 that used cross-species functional MRI data to sort autism into two connectivity-defined subtypes, one dominated by lower brain connectivity and the other by higher connectivity. The work combined scans from 20 genetic mouse models of autism risk with a multicenter human dataset. (nature.com) The researchers said the same two broad signatures appeared across species. The paper was published as autism research continues to grapple with why people given the same diagnosis can show very different biology and behavior. (lifescience.net) ### Which two subtypes did the researchers identify? The May 15 paper described a hypoconnectivity-dominant subtype and a hyperconnectivity-dominant subtype. (nature.com) In the authors’ account, the first showed reduced functional coupling across brain networks, while the second showed increased connectivity patterns. Nature’s summary of the article said Pagani and colleagues linked the lower-connectivity subtype to synaptic pathways and the higher-connectivity subtype to transcriptional and immune-related pathways. The paper’s abstract said those biological links first emerged in the mouse data and were then recapitulated in humans. (nature.com) ### How large was the human dataset? The human analysis included 940 people with idiopathic autism and 1,036 neurotypical individuals, according to the paper abstract indexed online. The dataset was multicenter, a detail the authors highlighted in describing the human replication. (lifescience.net) The 20 mouse models gave the team a way to compare connectivity patterns tied to different autism-risk genes. The cross-species design was central to the study’s claim that the two imaging signatures map onto distinct biology rather than reflecting only statistical clustering in one human cohort. (nature.com) ### Why did the authors use mice and humans in the same study? Nature Neuroscience said the study identified “two principal dysconnectivity signatures” across 20 mouse models of autism risk and then showed analogous patterns in autistic humans. The journal’s summary described the result as a translational framework for fMRI phenotyping. (lifescience.net) The authors argued that autism heterogeneity has long made it difficult to connect genes to brain dynamics. By starting with genetically defined mouse models and extending the analysis to people, they sought to test whether variation in brain connectivity reflects separable biological mechanisms. (nature.com) ### Does this mean autism can now be diagnosed with a brain scan? The paper did not say fMRI is ready to replace current autism diagnosis. The abstract said the work provides “a new empirical framework for targeted subtyping of the autism spectrum,” which is narrower than a claim of a clinical diagnostic test. (nature.com) Brain & Behavior Research Foundation, in a May 15 item about the study, said the result could help guide more targeted therapies. That language points to a possible research use for stratifying participants or studying mechanisms, not to an immediate change in clinical practice. (nature.com) ### What makes this different from earlier autism imaging work? Earlier autism imaging studies have often reported mixed findings, including both under-connectivity and over-connectivity, without a clear consensus across cohorts. A 2020 Human Brain Mapping paper, for example, reported nonreplication of some functional connectivity differences in autism across datasets. (lifescience.net) The new study’s claim is more specific: both patterns may be real, but concentrated in different subgroups. That framing is consistent with the paper’s central result that autism-related connectivity differences can be parsed into two dominant subtypes rather than averaged into a single inconsistent signal. (bbrfoundation.org) ### Who led the work, and what comes next? Nature’s 2026 article listing identified the paper as “Autism subtypes identified using cross-species functional connectivity analyses” by Marco Pagani, Valerio Zerbi, Silvia Gini and co-authors, with Adriana Di Martino and Alessandro Gozzi among the named researchers. (pubmed.ncbi.nlm.nih.gov) The article appeared in Nature Neuroscience on May 15. The next step described in the paper’s framing is targeted subtyping using fMRI features in future research cohorts. Any follow-on work will likely test whether those subtypes track treatment response or developmental course in named human cohorts beyond the 940 autistic participants already analyzed. (lifescience.net) (nature.com)