A foundational atlas of autism protein interactions reveals molecular convergence.

Despite the identification of a large number of high-confidence (hc) autism spectrum disorder (ASD) risk genes, translation to viable treatment targets remains elusive. Using affinity purification-mass spectrometry and HEK293T cells, we constructed a protein-protein interaction (PPI) network involving 100 hcASD risk genes, revealing over 1,800 PPIs, the vast majority of which have not been previously described. Importantly, the interactors are highly expressed in the human brain and neural progenitor cells, and enriched for ASD, but not schizophrenia, genetic risk. The network is highly connected, converges on protein complexes, including those involved in neurogenesis, tubulin biology, transcriptional regulation, and chromatin modification. We also generated a PPI map using over 50 patient-derived missense variants to identify differential physical interactions. Using these data, we leveraged AlphaFold2 predictions to prioritize direct PPIs as well as specific variants for interrogation in Xenopus tropicalis and human forebrain organoids. One such mutation in the transcription factor FOXP1, which disrupts interaction with FOXP4, led to a reconfiguration of DNA binding sites and disruption of neurogenesis, with a particularly strong impact on deep layer cortical and subplate neurons,consistent with a large body of work in ASD. Overall, this work offers critical new insights into molecular mechanisms underlying ASD that results from large effect genetic variants and describes a powerful platform to develop and test novel therapeutic strategies for a wide range of genetically-defined conditions.