Reading in data to TwoSampleMR options

examples.r

# MR-Base (i.e. using the cloud)

library(TwoSampleMR)
metab_ids <- c("met-e-1", "met-e-2")

# Extract data - from API
exp_dat <- extract_instruments("alz_id")
out_dat <- extract_outcome_data(exp_dat$SNP, metab_ids)

# Harmonise and run
dat <- harmonise_data(exp_dat, out_dat)
mr(dat)


# Running locally / on HPC (i.e. using GWAS VCF files on bc4)

library(gwasglue)
library(gwasvcf)
library(TwoSampleMR)

# Set default locations for bcftools, plink and ld reference files etc for use in bc4
set_bcftools()
set_plink()
set_bc4_files()

# extract data from vcfs using this helper function
# see documentation here https://mrcieu.github.io/gwasglue/reference/make_TwoSampleMR_dat.html
dat <- make_TwoSampleMR_dat(id1="alz_id", id2=metab_ids)

# run as usual
mr(dat)



# Running locally / on HPC (but using unprocessed txt.gzip files)

# Need to extract the SNPs you're interested in either in R using this function
# https://mrcieu.github.io/TwoSampleMR/reference/read_outcome_data.html
# Also provide arguments so it knows which columns to read in beta, se etc
exp_dat <- extract_instruments("alz_id")
out_dat1 <- read_outcome_data(filename="path/to/file1.txt.gz", snps=exp_dat$SNP)

# You can loop across all files by putting this in a for loop or whatever. It'll be quite slow though.

# Alternatively in bash do this sort of thing:

## make a snplist.txt that lists all the SNPs you want from the outcome metab datasets

# Create dir for outputs
mkdir -p /path/to/extract

# Create filelist.txt that lists all the filenames

# This is a loop that will go dataset by dataset extracting the SNPs you want
while IFS= read -r id
do
	# For every line in filelist, set that line to be a variable called ${id}
	zgrep -wf snplist.txt /path/to/gzipfiles/${id} | gzip -c > /path/to/extract/${id}
done < filelist.txt


# Back in R read all the data in and format to TwoSampleMR format using format_data