example harmonisation script for multiple GWAS summary datasets

harmonise.r

library(dplyr)

standardise <- function(d, ea_col="ea", oa_col="oa", beta_col="beta", eaf_col="eaf", chr_col="chr", pos_col="pos", vid_col="vid")
{
    toflip <- d[[ea_col]] > d[[oa_col]]
    d[[eaf_col]][toflip] <- 1 - d[[eaf_col]][toflip]
    d[[beta_col]][toflip] <- d[[beta_col]][toflip] * -1
    temp <- d[[oa_col]][toflip]
    d[[oa_col]][toflip] <- d[[ea_col]][toflip]
    d[[ea_col]][toflip] <- temp
    d[[vid_col]] <- paste0(d[[chr_col]], ":", d[[pos_col]], "_", d[[ea_col]], "_", d[[oa_col]])
    d
}

harmonise <- function(...)
{
    l <- list(...)
    
    # standardise each dataset and get new snp IDs
    l <- lapply(l, standardise)

    # get the SNPs in common across all datasets
    # arrange to be in the same order
    snpids <- Reduce(intersect, lapply(l, function(x) x$vid))
    lapply(l, function(x) {
        x %>% 
        filter(vid %in% snpids) %>%
        arrange(chr, pos, ea, oa)
    })
}

d1 <- tibble(
    snp=paste0("rs", 1:4),
    chr=10,
    pos=1:4,
    ea=c("C", "G", "A", "T"),
    oa=c("G", "A", "T", "G"),
    eaf=runif(4),
    beta=rnorm(4)
)
d1

d2 <- tibble(
    snp=paste0("rs", c(1:3, 3, 5)),
    chr=10,
    pos=c(1:3, 3,5),
    ea=c("G", "G", "A", "A", "T"),
    oa=c("C", "A", "T", "G", "G"),  # different alleles at rs3
    eaf=runif(5),
    beta=rnorm(5)
)
d2

d3 <- tibble(
    snp=paste0("rs", c(3,2,6)),
    chr=10,
    pos=c(3,2,6),
    ea=c("A", "G", "T"),
    oa=c("T", "A", "A"),
    eaf=runif(3),
    beta=rnorm(3)
)
d3

harmonise(d1, d2, d3)