Integrative Analyses of Autism and Intellectual Disability Exome Data Reveal Similarities and Divergences and Identify Novel Risk Genes for Both Disorders

Saturday, May 13, 2017: 1:27 PM
Yerba Buena 10-14 (Marriott Marquis Hotel)
S. De Rubeis1,2, A. E. Cicek3,4, L. Klei5, B. Devlin5, J. D. Buxbaum2,6 and K. Roeder7,8, (1)Department of Psychiatry, Icahn School of Medicine at Mount Sinai, New York, NY, (2)Seaver Autism Center for Research and Treatment, Icahn School of Medicine at Mount Sinai, New York, NY, (3)Department of Computer Engineering, Bilkent University, Ankara, TURKEY, (4)Computational Biology Department, Carnegie Mellon University, Pittsburgh, PA, (5)Department of Psychiatry, University of Pittsburgh, Pittsburgh, PA, (6)Seaver Autism Center for Research and Treatment, Mount Sinai School of Medicine , New York, NY, (7)Department of Psychiatry, Carnegie Mellon University, Pittsburgh, PA, (8)Department of Statistics, Carnegie Mellon University, Pittsburgh, PA

Autism spectrum disorder (ASD) and intellectual disability (ID) are known to have a complex genetic architecture. ASD and ID frequently co-occur and recent analyses suggest that the genetic architecture might differ in individuals with low or high cognitive function. These findings raise questions about shared risk and compel analyses that intersect the genetic studies in ASD and ID since they have so far proceeded along parallel routes.


The objective was to understand the nature of shared genetic risk between ASD and ID.


We assimilated rare variants identified via exome sequencing of 4,216 ASD and 1,479 ID trios, as well as 869 ASD cases and 2,829 control samples. We assessed the evidence for association using TADA, a statistical algorithm that combines information across different classes of evidence and produces a set of dominant-effect risk genes with varying levels of significance, measured as q-values. Genes were categorized based on level of evidence as strong (q < 0.05), moderate (0.05 < q < 0.3), weak (0.3 < q < 0.6) and negligible (q > 0.6).


For ASD and ID we find 31 and 64 genes with strong evidence (referred to as tASD and tID), 92 and 99 genes with moderate evidence, and 355 and 320 genes with weak evidence of association, respectively. A cross-classification of genes by FDR level shows substantial overlap in these risk gene lists (Chi-square p < 10-15): 12 genes show strong evidence (p < 10-20) (referred to as tASD.ID genes) of conferring risk for both disorders. Based on the pattern of de novo LoF, we estimated the total number of autosomal dominant genes in which de novo LoF imparts substantial risk for ASD (KASD) and ID (KID), respectively. We estimated a 95% confidence interval for KASD of 500-950 and KID of 185-225. Impact on nonverbal IQ (NVIQ) in probands diagnosed with ASD differs across categories: a de novo loss-of-function (LoF) mutation in a tID or tASD.ID gene had an average drop on NVIQ of about 24 points, while a de novo LoF mutation in a tASD gene had a weaker effect (~12 points reduction). The substantial genetic overlap in signal for ID and ASD motivates pooling the data to enhance power and thus gene discovery. Using this strategy, we identified 16 additional risk genes with q < 0.05, many of which cross-validated in complementary datasets.


Despite having a much larger sample of ASD probands, we identify twice as many ID-related genes. This is consistent with a model in which de novo mutations contribute risk in a higher proportion of ID subjects than ASD and/or each mutation contributes a greater degree of risk in ID than ASD. We identified 12 genes that confer risk to both disorders and observed a dearth of genes that show strong evidence for association with ID but no evidence of association with ASD and vice versa. Finally, mutations in ID genes can have a marked impact on IQ of ASD subjects, whereas mutations in ASD-only genes have smaller impact.

See more of: Gene Discovery in ASD
See more of: Genetics