Beyond SNP Heritability: Polygenicity and Discoverability of Phenotypes Estimated with a Univariate Gaussian Mixture Model

Abstract

Of signal interest in the genetics of human traits is estimating their polygenicity (the proportion of causally associated single nucleotide polymorphisms (SNPs)) and the discoverability (or effect size variance) of the causal SNPs. Narrow-sense heritability is proportional to the product of these quantities. We present a basic model, using detailed linkage disequilibrium structure from an extensive reference panel, to estimate these quantities from genome-wide association studies (GWAS) summary statistics for SNPs with minor allele frequency >1%. We apply the model to diverse phenotypes and validate the implementation with simulations. We find model polygenicities ranging from ≃ 2 × 10−5 to ≃ 4 × 10−3, with discoverabilities similarly ranging over two orders of magnitude. A power analysis allows us to estimate the proportions of phenotypic variance explained additively by causal SNPs at current sample sizes, and map out sample sizes required to explain larger portions of additive SNP heritability. The model also allows for estimating residual inflation.

Making the MOSTest of imaging genetics

“…As an example, Figure 3 shows the maps for the top two hits (rs1080066 on chromosome 15, p=1.2x10 -305 , and rs13107325 on chromosome 4, p=3.1x10 -124 ), all other maps are available in the Supplementary Material. These maps revealed anterior-posterior gradients as well as hemisphere-specific effects of some of the lead SNPs, in line with reported genetic patterns of the brain 15,16 .…”

ENIGMA and global neuroscience: A decade of large-scale studies of the brain in health and disease across more than 40 countries

“…Current ENIGMA sample sizes (which now exceed 50,000) are sufficiently large to identify genetic associations at a pace comparable to that of GWAS for other phenotypes. In a recent analysis, Holland 35 contrasted rates of discovery of genetic loci by ENIGMA and the PGC and noted the distribution of effect sizes for some brain measures (e.g., putamen volume) may indeed be enriched for slightly larger effects compared to behavioral traits (see also Le and Stein 36 and Franke et al 37 ). Still, a central understanding gained from the ENIGMA association screens is that neuroimaging genetics studies-just like analyses of behavioral measures, require tens (perhaps hundreds) of thousands of participants to obtain robust and reproducible effects of common polymorphisms.…”

Quantifying the polygenic architecture of the human cerebral cortex: Extensive genetic overlap between cortical thickness and surface area

“…We applied causal mixture models 3,9 to the GWAS summary statistics, using the MiXer tool (https://github.com/precimed/mixer). For each SNP, , univariate MiXeR models its additive genetic effect of allele substitution, * , as a point-normal mixture, * = (1 − 0 ) (0,0) + 0 (0, " # ), where 0 represents the proportion of non-null SNPs (`polygenicity`) and " # represents variance of effect sizes of non-null SNPs (`discoverability`).…”

“…Despite largescale efforts, the majority of these genetic variants remains unknown 1 . This is in part due to the genetic signal of cortical morphology being distributed across many causal variants, each having a small effect 2,3 . Our ability to identify causal SNPs can be improved not only by increasing sample sizes to boost statistical power, but also by employing better delineated, less noisy brain measures that better map onto the biology we seek to understand.…”