9 Genetic Correlations

Load R Packages

library(tidyverse)    # Data wrangling 
library(GenomicSEM)

Methods

Tools & Publications

R, GenomicSEM, LDSC, HDL

Bulik-Sullivan, B. et al. An atlas of genetic correlations across human diseases and traits. Nat Genet 47, 1236–1241 (2015).
Ning, Z., Pawitan, Y. & Shen, X. High-definition likelihood inference of genetic correlations across human complex traits. Nat Genet 52, 859–864 (2020).

Genetic correlation (rg) refers to the degree to which the genetic determinants of two traits overlap - the proportion of variance that two traits share due to genetic causes. A positive genetic correlation between two traits implies that the same genetic variants are influencing both traits in the same direction. Conversely, a negative genetic correlation implies that the genetic variants influencing one trait are having the opposite effect on the other trait.

LDSC: Linkage disequilibrium score regression (LDSC) leverages linkage disequilibrium (LD), the non-random association of alleles at different loci, to estimate genetic correlations between two traits. This method operates on the premise that single nucleotide polymorphisms (SNPs) with a higher count of LD partners (thus having a higher LD score) are typically more associated with a trait due to polygenicity, a condition where numerous genetic variants each exert a minor effect.

HDL: High-definition likelihood (HDL) provides genetic correlation estimates that have higher accuracy and precision compared to LDSC. HDL achives this by using a full likelihood-based method that leverages LD information across the whole genome, where as LDSC only use partial information.

Munge GWAS SumStats

## Summary statistics 
Willer2013ldl = "resources/Willer2013ldl.chrall.CPRA_b37.tsv.gz"
Graham2021ldl = "resources/Graham2021ldl.chrall.CPRA_b37.tsv.gz"
KunkleAD = "resources/Kunkle2019load_stage123.chrall.CPRA_b37.tsv.gz"
BellenguezAD = "resources/Bellenguez2022load.chrall.CPRA_b37.tsv.gz"

## LD Structure 
ld_path = "resources/eur_w_ld_chr/"

## HAPMAP3 SNPs
hm3_path = "resources/w_hm3.snplist"


GenomicSEM::munge(
  files = c(Willer2013ldl, Graham2021ldl, KunkleAD, BellenguezAD), 
  hm3 = hm3_path, 
  trait.names = c("Willer2013ldl", "Graham2021ldl", "KunkleAD", "BellenguezAD"), 
  maf.filter = 0.05, 
  column.names = list(
    SNP='DBSNP_ID', 
    MAF='AF', 
    A1='ALT',
    A2='REF', 
    effect='BETA', 
    N = "N"
  ), 
  overwrite=FALSE
)

LDSC

ldsc.covstruct <- GenomicSEM::ldsc(
     traits = c("Willer2013ldl.sumstats.gz", "Graham2021ldl.sumstats.gz", "BellenguezAD.sumstats.gz", "KunkleAD.sumstats.gz"),
     trait.names = c("Willer2013ldl", "Graham2021ldl", "BellenguezAD", "KunkleAD"), 
     sample.prev = c(NA, NA, 0.18, 0.37),
     population.prev = c(NA, NA, 0.31, 0.31),
     ld = ld_path, 
     wld = ld_path,
     stand = TRUE
     )

HDL

hdl.covstruct <- GenomicSEM::hdl(
     traits = c("Willer2013ldl.sumstats.gz", "Graham2021ldl.sumstats.gz", "BellenguezAD.sumstats.gz", "KunkleAD.sumstats.gz"),
     trait.names = c("Willer2013ldl", "Graham2021ldl", "BellenguezAD", "KunkleAD"), 
     sample.prev = c(NA, NA, 0.18, 0.37),
     population.prev = c(NA, NA, 0.31, 0.31),
     LD.path="resources/UKB_imputed_hapmap2_SVD_eigen99_extraction/", 
     method = "piecewise"
     )