LD Score regression distinguishes confounding from polygenicity in genome-wide association studies

Abstract

Both polygenicity (i.e., many small genetic effects) and confounding biases, such as cryptic relatedness and population stratification, can yield an inflated distribution of test statistics in genome-wide association studies (GWAS). However, current methods cannot distinguish between inflation from true polygenic signal and bias. We have developed an approach, LD Score regression, that quantifies the contribution of each by examining the relationship between test statistics and linkage disequilibrium (LD). The LD Score regression intercept can be used to estimate a more powerful and accurate correction factor than genomic control. We find strong evidence that polygenicity accounts for the majority of test statistic inflation in many GWAS of large sample size.

Polygenic risk scores for schizophrenia and bipolar disorder associate with addiction

“…Effects and odds ratios (ORs) are reported for one unit change in polygenic score. To account for inflation due to relatedness and population stratification, the test statistics for each analysis was divided by an inflation factor estimated from LD score regression (Bulik‐Sullivan et al 2015b). …”

“…In particular, the PRS methodology can be applied to subjects who do not have the psychiatric disorders, providing separation of the effect of the diseases themselves from those of the underlying genetic risk factors. PRSs and LD score regression (Bulik‐Sullivan et al 2015b) have revealed shared genetic etiology between SCZ and BPD (Bulik‐Sullivan et al 2015a), between psychosis and cannabis use (Power et al 2014), and use of other substances (Bulik‐Sullivan et al 2015a). Given these relationships, we hypothesized that the SCZ‐ and BPD‐PRSs could be used to reveal the extent of genetic overlap between these psychotic disorders and substance use disorders.…”

Investigating shared aetiology between type 2 diabetes and major depressive disorder in a population based cohort

“…This method uses the correlational nature of SNPs such that SNPs with high LD will have higher average χ 2 statistics than those with low LD. To estimate genetic correlations the product of two z‐scores from GWAS of two traits can be regressed onto the LD score and the slope of the regression used to estimate genetic covariance [Bulik‐Sullivan et al, 2015]. The intercept was left unconstrained as the degree of sample overlap between DIAGRAM T2D and PGC MDD cohorts was unknown.…”

“…One limitation of using PRS to explore causal relationships is the risk of pleiotropy: associations may arise from causality or pleiotropy, particularly when thousands of SNPs comprise the PRS. Another technique, linkage disequilibrium (LD) score regression uses the LD information from SNPs to compute genetic correlations between traits of interest from GWAS summary statistics [Bulik‐Sullivan et al, 2015]. This method is typically better powered to detect genetic correlations compared to PRS and provides more reliable estimates of the magnitude of genetic overlap between traits.…”

Using partitioned heritability methods to explore genetic architecture

“…Second, computationally efficient, partitioned LD score regression (LDSC) [11][12][13] relates the LD tagging of common SNPs to their GWAS summary statistics to estimate heritability. By quantifying the heritability enrichment of particular annotations (and not limiting to coding variation), partitioned LDSC can elucidate functional domains that contribute to phenotypic variation, illuminating differences in the genetic architecture of traits, such as levels of triglycerides and lowdensity lipoprotein (LDL) 13 , and evolutionary connections, such as purifying selection and allele age 11 .…”