Project Summary
Autism spectrum disorder (ASD) is common and often disabling. Currently, there are no pharmacologic
treatments for its core symptoms: impairment in social communication and restricted/repetitive behavior. The
high heritability of ASD suggests that understanding its genetic risk factors will help define ASD
pathophysiology and facilitate therapeutic development. The largest genetic risk factor for ASD on a population
level is common polygenic variation, the additive influence of thousands of small effect SNPs estimated with
GWAS. However, interpretation of polygenic risk is challenging: most disease-associated SNPs are non-
coding, the identity of their target gene(s) is uncertain, and their influence on gene expression is often
unknown. Thus, the genes and biological pathways affected by common polygenic risk for ASD remain poorly
defined. In contrast, recent studies of rare coding variation have associated >70 genes with ASD (“ASD
Genes”), which collectively highlight ASD-relevant biological processes. To define the biology of polygenic risk
for ASD, it is critically important to understand the degree to which polygenic risk converges with the genes
and pathways identified through rare variant studies. We propose to apply three novel statistical and
functional approaches to (a) interpret common polygenic risk for ASD and (b) assess the convergence
of common polygenic risk for ASD and ASD Genes. In Aim 1, we introduce the Stratified-polygenic
Transmission Disequilibrium Test (S-pTDT), which is powered to examine polygenic risk arising from small
regions of the genome. We will use S-pTDT to quantify the concentration of polygenic risk for ASD in and
around ASD Genes. In Aim 2, we introduce an approach to understand how polygenic risk influences gene
expression, leveraging a large single-nucleus transcriptomic dataset of adult human brain tissue. We will use
this approach to test the hypothesis that polygenic risk for ASD decreases expression of ASD Genes, and to
nominate candidate causal genes in genomic regions with heightened polygenic relevance in ASD. In Aim 3,
we introduce the Abstract Mediation Model (AMM), which quantifies the proportion of common-variant
heritability mediated by gene sets while accounting for uncertainty in the gene targets of non-coding SNPs. We
will apply AMM to estimate the extent to which polygenic risk for ASD is mediated by ASD Genes and other
ASD-relevant gene sets. These Aims will improve our understanding of the genetic underpinnings of ASD,
strengthen biological and therapeutic hypotheses, and generate methods applicable to other neuropsychiatric
conditions. Finally, this research training will occur in the stimulating environment of the Harvard/MIT MD/PhD
Program and Broad Institute of MIT and Harvard. Under the guidance of his sponsors, the fellowship PI will
complete five training goals that will create the foundation for a career as an NIH-funded physician-scientist.