An Expanded View of Complex Traits: From Polygenic to Omnigenic

Abstract

A central goal of genetics is to understand the links between genetic variation and disease. Intuitively, one might expect disease-causing variants to cluster into key pathways that drive disease etiology. But for complex traits, association signals tend to be spread across most of the genome-including near many genes without an obvious connection to disease. We propose that gene regulatory networks are sufficiently interconnected such that all genes expressed in disease-relevant cells are liable to affect the functions of core disease-related genes and that most heritability can be explained by effects on genes outside core pathways. We refer to this hypothesis as an "omnigenic" model.

Figure 1

Genome-wide signals of association with height. A. Genome-wide inflation of small p-values from the GWAS for height, with particular enrichment among expression quantitative trait loci and single nucleotide polymorphisms (SNPs) in active chromatin (H3K27ac). B. Estimated fraction of SNPs associated with non-zero effects on height (Stephens, 2016), as a function of Linkage Disequilibrium Score (i.e., the effective number of SNPs tagged by each SNP (Bulik-Sullivan et al., 2015b)). Overall we estimate that 62% of all SNPs are associated with a non-zero effect on height. The best-fit line estimates that 3.8% of SNPs have causal effects. C. Estimated mean effect size for SNPs, sorted by GIANT p-value with the direction (sign) of effect ascertained by GIANT. Replication effect sizes were estimated using data from the Health and Retirement Study (HRS). The points show averages of 1,000 consecutive SNPS in the p-value sorted list. The effect size on the median SNP in the genome is about 10% of that for genome-wide significant hits.

Figure 2

Heritability tends to be enriched in regions that are transcriptionally active in relevant tissues. (a) Contributions to heritability (relative to random SNPs) as a function of chromatin context. There is enrichment for signal among SNPs that are in chromatin active in the relevant tissue, regardless of the overall tissue breadth of activity. (b) Genes with brain-specific expression show the strongest enrichment of schizophrenia signal (left), but broadly expressed genes contribute more to total heritability due to their greater number (right).

Figure 3

Gene ontology enrichments for three diseases, with categories of particular interest labeled. The x-axis indicates the fraction of genes in each category; the y-axis shows the fraction of heritability assigned to each category relative to all other SNPs. Note that the diagonal indicates the genome-wide average across all SNPs; most GO categories lie above the line due to the general enrichment of signal in and around genes. Analysis by stratified LD score regression (Finucane et al., 2015).

Figure 4

An omnigenic model of complex traits. (A) For any given disease phenotype, a limited number of genes have direct effects on disease risk. However by the small world property of networks, most expressed genes are only a few steps from the nearest core gene, and thus may have non-zero effects on disease. Since core genes only constitute a tiny fraction of all genes, most heritability comes from genes with indirect effects. (B) Diseases are generally associated with dysfunction of specific tissues; genetic variants are only relevant if they perturb gene expression (and hence network state) in those tissues. For traits that are mediated through multiple cell types or tissues, the overall effect size of any given SNP would be a weighted average of its effects in each cell type.