mr-a-b

mr-a-b.r

library(dplyr)
library(data.table)
library(TwoSampleMR)

bmi_id <- "ukb-b-19953"

chd_id <- "ukb-b-3983"

datadir <- "/mnt/storage/private/mrcieu/research/mr-eve/UKBB_replication/replication/results"


bmiA <- fread(file.path(datadir, bmi_id, "discovery.statsfile.txt.gz"), header=TRUE)
bmiB <- fread(file.path(datadir, bmi_id, "replication.statsfile.txt.gz"), header=TRUE)

chdA <- fread(file.path(datadir, chd_id, "discovery.statsfile.txt.gz"), header=TRUE)
chdB <- fread(file.path(datadir, chd_id, "replication.statsfile.txt.gz"), header=TRUE)


cor(bmiA$BP, bmiB$BP)
cor(bmiA$BETA, bmiB$BETA)

igddir <- "/mnt/storage/private/mrcieu/research/scratch/IGD"
bmi_clumped <- scan(file.path(igddir, "data", "public", bmi_id, "clump.txt"), what="character")

instrumentsA <- subset(bmiA, P_BOLT_LMM_INF < 5e-8 & SNP %in% bmi_clumped)

instrumentsB <- subset(bmiB, P_BOLT_LMM_INF < 5e-8 & SNP %in% bmi_clumped)

table(instrumentsA$SNP %in% instrumentsB$SNP)
table(instrumentsB$SNP %in% instrumentsA$SNP)


instrumentsA_rep <- subset(bmiB, SNP %in% instrumentsA$SNP)
instrumentsB_rep <- subset(bmiA, SNP %in% instrumentsB$SNP)
outcomeB_A <- subset(chdB, SNP %in% instrumentsA$SNP)
outcomeB_B <- subset(chdB, SNP %in% instrumentsB$SNP)
outcomeA_A <- subset(chdA, SNP %in% instrumentsA$SNP)
outcomeA_B <- subset(chdA, SNP %in% instrumentsB$SNP)
all(instrumentsA_rep$SNP == outcomeB_A$SNP)
all(instrumentsA_rep$ALLELE1 == outcomeB_A$ALLELE1)
all(instrumentsA_rep$SNP == outcomeB_A$SNP)
all(instrumentsA_rep$ALLELE1 == outcomeB_A$ALLELE1)


l <- list()

what="exposure_discovery A, exposure_replication B, outcome B"
dat <- merge(instrumentsA_rep, outcomeB_A, by="SNP")
l[[1]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)


what <- "exposure_discovery A, outcome A"
dat <- merge(instrumentsA, outcomeB_A, by="SNP")
l[[2]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)


what <- "exposure_discovery B, exposure_replication A, outcome A"
dat <- merge(instrumentsB_rep, outcomeA_B, by="SNP")
l[[3]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)


what <- "exposure_discovery B, outcome B"
dat <- merge(instrumentsB, outcomeB_B, by="SNP")
l[[4]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)


what <- "exposure_discovery B, outcome B"
dat <- merge(instrumentsB, outcomeA_B, by="SNP")
l[[5]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)


what <- "bmi instrument replication B"
dat <- merge(instrumentsB, instrumentsB_rep, by="SNP")
l[[6]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)

what <- "bmi instrument replication A"
dat <- merge(instrumentsA, instrumentsA_rep, by="SNP")
l[[7]] <- mr_ivw(
	b_exp=dat$BETA.x, 
	b_out=dat$BETA.y, 
	se_exp=dat$SE.x, 
	se_out=dat$SE.y) %>% 
	as.data.frame %>% mutate(what=what)

res <- bind_rows(l)