Polygenic Risk for Autism Spectrum Disorder in the Study to Explore Early Development

Panel Presentation
Saturday, May 4, 2019: 10:30 AM
Room: 517B (Palais des congres de Montreal)
K. S. Benke1, C. Ladd-Acosta2, C. J. Newschaffer3, M. D. Fallin2, L. A. Croen4, L. Schieve5, J. Daniels6, A. M. Reynolds7 and D. Schendel8, (1)Mental Health, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, (2)Wendy Klag Center for Autism and Developmental Disabilities, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, (3)AJ Drexel Autism Institute, Philadelphia, PA, (4)Division of Research, Kaiser Permanente, Oakland, CA, (5)Centers for Disease Control and Prevention, Atlanta, GA, (6)University of North Carolina at Chapel Hill, Chapel Hill, NC, (7)University of Colorado Denver School of Medicine, Aurora, CO, (8)Aarhus University, Aarhus, Denmark

Autism spectrum disorder (ASD) is a polygenic disease that is moderately to highly heritable. Genome-wide association studies (GWAS) have been successful in identifying common genetic variants that predispose to complex traits.

The largest ASD GWAS meta-analysis to date includes >9 million single-nucleotide-polymorphisms (SNPs) in 18,381 ASD cases and 27,969 controls from the Psychiatric Genomic Consortium (PGC) and iPSYCH projects. These results provide, precise SNP-based association effect sizes that allow derivation of a polygenic risk score (PRS) to represent genetic susceptibility to ASD (ASD-PRS).


We applied GWAS effect sizes from the recent PGC/iPSYCH results to SNP data from the independent US population-based Study to Explore Early Development (SEED) to derive ASD-PRS and examine associations between ASD-PRS and ASD classification.


SEED is a multi-site case-control study of children aged 3-5 years with ASD and a control group drawn from the general population, born between September 2003 and August 2006. Genotyping was performed on either Illumina or Affymetrix arrays and imputed using SHAPEIT and IMPUTE2. ASD- PRS was generated via pruning, thresholding and scoring in PLINK. The same process was used to derive a psychiatric cross-disorder PRS using PGC results (http://www.med.unc.edu/pgc/results-and-downloads). All ASD-PRS associations with ASD were assessed by logistic regression, accounting for sex and five principal components of genetic ancestry.


The PRS analysis included 1302 SEED children: 861 boys (467 with ASD) and 441 girls (115 with ASD). After evaluating Nagelkerke’s R-squared statistic across a range of discovery p-value thresholds, we observed that pdiscovery=0.30 maximized the variance explained. All PRS associations reflect this cutoff for SNPs to be selected into the summed score. The odds ratio of ASD for a single standard deviation increase in the ASD-PRS score, adjusted for sex and genetic ancestry, was 1.42 (95%CI: 1.13, 1.78; p-value=0.0028), which explained 0.87% of the variance in ASD status. A similar finding was observed for a subset of 799 children with European ancestry (OR=1.49; 95%CI: 1.12, 2.00; p-value=0.0066), which explained 1.1% of the variance in ASD status. We also observed an association with the CrossDisorder-PRS in the full sample (OR=1.28, 95%CI: 1.09,1.49, pl=0.0021) and in children of European descent (OR=1.46, 95%CI: 1.17, 1.82, p=0.0008). In a model that includes both ASD-PRS and CrossDisorder-PRS, the odds ratio for ASD-PRS, adjusted for the CrossDisorder-PRS, was 1.43 (95%CI: 1.07, 1.92, p=0.015), and the odds ratio for CrossDisorder-PRS, adjusted for ASD-PRS, was 1.43 (95%CI: 1.14, 1.79, p=0.0017).


To our knowledge, this is the first effort to leverage the PGC/iPSYCH GWAS discovery results to investigate the utility of an ASD-PRS in a population-based ASD case-control sample. We show a strong main effect indicating genetic susceptibility for ASD, in a multi-ethnic sample and stratified by European ancestry. The variance explained, though small (~1%), is similar to PGC-based PRS for other psychiatric conditions (i.e., schizophrenia). We additionally show that the ASD-PRS association persists after adjustment for a general genetic susceptibility to psychiatric disorders. Further work will compare these results to similar analyses using a non-psychiatric phenotype to confirm the specificity of this finding.