Thread by @amandambuch

1/12: Check out our new @NatureNeuro paper! Molecular and network-level mechanisms explaining individual differences in #autism spectrum disorder by @amandambuch with Logan Grosenick and Conor Liston @WeillCornell @nyphospital @WCMPsychiatry doi.org/10.1038/s41593…

https://twitter.com/NatureNeuro/status/1633916931672621058

2/12: #Biomarkers have transformed diagnosis and treatment in medicine, but are elusive in neuropsychiatric syndromes such as autism spectrum disorder (ASD), a heterogeneous neurodevelopmental disorder.

3/12: Each individual with ASD presents with a different combination of strengths and deficits including a range of difficulties in social interaction/communication, repetitive/ritualistic behaviors, differing levels of intellectual disability and various medical comorbidities.

4/12: #Autism is highly heritable, but rarely linked to just one gene. Instead, it is associated with genetic variation in 100s of genes (it is highly polygenic). This phenotypic heterogeneity is a major obstacle to defining the mechanisms of ASD, and to discovering new therapies

5/12: To investigate this, we asked, “Does polygenic variation in autism spectrum disorder manifest as differences in brain connectivity and behavior that define different ASD subgroups?” doi.org/10.1038/s41593…

6/12: First, we used #MachineLearning of functional brain connectivity (resting state #fMRI) and measures of atypical behaviors to reveal 3 latent brain-behavior dimensions (distinct brain connectivity patterns associated with ASD-related behaviors not directly observable)

7/12: Second, we performed hierarchical clustering on these 3 latent brain-behavior dimensions, and identified 4 distinct #autism subgroups. These were highly robust to data permutation, and replicated in out-of-sample data from a completely different study of ASD individuals.

8/12: Each of the four autism subgroups had distinct patterns of atypical neurobiology that correlated with distinct patterns of autism-related behaviors.

9/12: Third, we found that the expression of different gene sets across the human brain predicted subgroup atypical connectivity, and was enriched for ASD-related, synaptic/immune signaling, and protein translation gene sets to differing degrees in each ASD subgroup.

10/12: Fourth, we showed that the highly ranked genes in each subgroup may modulate atypical brain connectivity in #autism through protein-protein interactions in gene regulatory networks known to be involved in ASD.

11/12: Finally, we used text-mining of published research to ask whether candidate hub genes important in each subgroup were associated with different #autism symptoms in the literature. This independently confirmed that our subgroups were related to behavioral effect keywords.

12/12: Overall, we found that stable ASD-related brain connectivity patterns subserve different ASD subgroups and map to distinct molecular processes and clinical behaviors, yielding testable hypotheses of brain network and genetic substrates underlying phenotypic heterogeneity.