example of causal effect of all CpGs on outcome

cpg_outcome_scan.R

# Load libraries
library(TwoSampleMR)
library(MRInstruments)

# Load the ARIES mQTL catalog
data(aries_mqtl)

# Get the mQTLs from just one time point
cpglist <- subset(aries_mqtl, age == "Childhood")

# Remove duplicated SNPs
cpglist <- subset(cpglist, !duplicated(SNP))

# Convert them into the format required for 2-sample MR
exposure_dat <- format_aries_mqtl(cpglist)

# Find the id of the outcome of interest
ao <- available_outcomes()

# e.g. BMI = 2

# Get the CpG SNPs from the outcome
# From this point it could be very slow with 40k CpGs - might be better to split it up into smaller chunks
out <- extract_outcome_data(ex$SNP, c(2))

# Harmonise the exposure and outcome data
dat <- harmonise_data(ex, out)

# Perform MR of each CpG on outcome data
res <- mr(dat, method_list=c("mr_meta_fixed"))

# Save the results
save(cpglist, out, dat, res, file="cpg_results.RData")

# Q-Q plot
library(GenABEL)
P <- subset(res, outcome == i)$pval
l <- estlambda(P)
j <- strsplit(i, split="\\|")[[1]][1]
nom <- paste(j, "\n", "number of tests = ", length(P), "\nlambda = ", round(l$estimate, 3), ", se = ", round(l$se, 3), sep="")
qq(P, main=nom)