Parsing Autism Heterogeneity: Transcriptomic Subgrouping of Imaging-Derived Phenotypes in Autism.

Abstract

Background: Neurodevelopmental conditions, such as autism, are highly heterogeneous both at the mechanistic and phenotypic level. Parsing heterogeneity is therefore vital for uncovering underlying processes that could inform the development of targeted, personalized support. The study aimed to parse heterogeneity in autism by identifying subgroups that converge at both phenotypic and molecular levels.

Methods: An imaging-transcriptomics approach was used to link neuroanatomical imaging-derived phenotypes in autism to whole-brain gene expression signatures provided by the Allen Human Brain Atlas. Neuroimaging and clinical data of N=359 autistic participants aged 6-30 years were provided by the EU-AIMS Longitudinal European Autism Project. Individuals were stratified using data-driven clustering techniques based on the correlation between brain phenotypes and transcriptomic profiles. The resulting subgroups were characterized on the clinical, neuroanatomical, and molecular level.

Results: We identified three subgroups of autistic individuals based on the correlation between imaging-derived phenotypes and transcriptomic profiles which showed different clinical phenotypes. The individuals with the strongest transcriptomic associations to imaging-derived phenotypes showed the lowest level of symptom severity. The genesets most characteristic for each subgroup were significantly enriched for genes previously implicated in autism etiology, including processes like synaptic transmission and neuronal communication, and mapped onto different gene ontology categories.

Conclusion: Autistic individuals can be sub-grouped based on the transcriptomic signatures associated with their neuroanatomical fingerprints, revealing subgroups that show differences in clinical measures. The study presents an analytical framework for linking neurodevelopmental and clinical diversity in autism to underlying molecular mechanisms, thus highlighting the need for personalized support strategies.