3 LD matrix

3.1 Data

# pQTLs
pqtls <- fread(
  "./data/00_data_pqtls.tsv",
  header = TRUE, data.table = FALSE, sep = "\t"
)

# LD matrix
load("./data/01_data_ldmat.Rda")

3.2 LD matrix

Sclerostin pQTLs
Sclerostin pQTLs and CD300LG variants

# LD matrix for pQTLs
ld_mat_pqtls <- ld_mat[
  match(pqtls$rsid, rownames(ld_mat)),
  match(pqtls$rsid, colnames(ld_mat)),
  drop = FALSE
]

# Compute r2 from r
ld_mat_pqtls <- round(ld_mat_pqtls^2, 3)

# Heatmap
heatmaply::heatmaply(
  ld_mat_pqtls,
  label_names = c("Variant 1", "Variant 2", "r2"),
  row_dend_left = TRUE,
  key.title = "r2"
)

Figure 3.1: LD matrix sclerostin pQTLs

Since CD300LG located nearby to SOST is known to be associated with HDL cholesterol and triglycerides, we examined whether any of the sclerostin pQTLs are in LD with either rs72836561 (a functional variant in CD300LG known to be associated with HDL cholesterol and triglycerides (Surakka et al. 2015)) and / or rs72836567 (the top associated variant with CD300LG expression in GTEx (GTEx Consortium 2020)). Two of the sclerostin pQTLs (rs1107747 & rs4793023) are mildly correlated (\(r^2\) > 0.05) with rs72836567. So, for these sclerostin pQTLs we used rs72836567 to adjust out any residual association with HDL cholesterol and triglycerides due to the effects of the CD300LG gene (Yang et al. 2012).

# LD matrix for pQTLs and CD300LG variants
ld_mat_pqtls_cd300lg <- ld_mat[
  match(c(pqtls$rsid, "rs72836561", "rs72836567"), rownames(ld_mat)),
  match(c(pqtls$rsid, "rs72836561", "rs72836567"), colnames(ld_mat)),
  drop = FALSE
]

# Compute r2 from r
ld_mat_pqtls_cd300lg <- round(ld_mat_pqtls_cd300lg^2, 3)

# CD300LG correlations
ld_mat_pqtls_cd300lg <- ld_mat_pqtls_cd300lg[
  pqtls$rsid,
  c("rs72836561", "rs72836567"),
  drop = FALSE
]

# Heatmap
heatmaply::heatmaply(
  ld_mat_pqtls_cd300lg,
  dendrogram = "none",
  label_names = c("Sclerostin pQTL", "CD300LG variant", "r2"),
  limits = c(0, 1),
  key.title = "r2"
)

Figure 3.2: LD matrix sclerostin pQTLs and CD300LG variants