PT  - JOURNAL ARTICLE
AU  - Luke J. O’Connor
AU  - Alkes L. Price
TI  - Distinguishing genetic correlation from causation across 52 diseases and complex traits
AID  - 10.1101/205435
DP  - 2017 Jan 01
TA  - bioRxiv
PG  - 205435
4099  - http://biorxiv.org/content/early/2017/10/24/205435.short
4100  - http://biorxiv.org/content/early/2017/10/24/205435.full
AB  - Mendelian randomization (MR) is widely used to identify causal relationships among heritable traits, but it can be confounded by genetic correlations reflecting shared etiology. We propose a model in which a latent causal variable mediates the genetic correlation between two traits. Under the LCV model, trait 1 is fully genetically causal for trait 2 if it is perfectly genetically correlated with the latent causal variable, and partially genetically causal for trait 2 if the latent variable has a higher genetic correlation with trait 1 than with trait 2. To quantify the degree of partial genetic causality, we define the genetic causality proportion (gcp), enabling us to describe genetically causal relationships non-dichotomously. We fit this model using mixed fourth moments and of marginal effect sizes for each trait, exploiting the fact that if trait 1 is causal for trait 2 then SNPs with large effects on trait 1 will have correlated effects on trait 2, but not vice versa. We performed simulations under a wide range of genetic architectures and determined that LCV, unlike state-of-the-art MR methods, produced well-calibrated false positive rates and reliable gcp estimates in the presence of genome-wide genetic correlations and asymmetric genetic architectures. We applied LCV to GWAS summary statistics for 52 traits (average N=326k), identifying fully or partially genetically causal effects (1% FDR) for 63 pairs of traits. Results consistent with the published literature included causal effects on myocardial infarction (MI) for LDL, triglycerides and BMI. Novel findings included an effect of LDL on bone mineral density, consistent with clinical trials of statins in osteoporosis. Our results demonstrate that it is possible to distinguish between genetic correlation and causation using genetic data.