In theboocock/coco: Correlation-based conditional analysis

    library(data.table)
    f_read = as.data.frame(fread("data2/kottgen_effect_filtered.txt", header=T))
    ld_matrix = as.matrix(fread("data2/slc2a9_only.ld", header=F))
    snpdat = as.data.frame(fread("data2/ABCG2.snpdat", header=T))
    idx_in = which(snpdat$RSID %in% f_read$MarkerName)
    ld_abcg2 = ld_matrix[idx_in, idx_in]
    snpdat_abcg2 = snpdat[idx_in,]
    f_merge = merge(f_read, snpdat_abcg2, by.x="MarkerName", by.y="RSID")
    f_merge = f_merge[order(f_merge$POS),]
    f_merge$Effect = ifelse(toupper(f_merge$Allele1) == toupper(f_merge$A2), f_merge$Effect, -f_merge$Effect)
    colnames(f_merge)[1] = "SNP"
    library(coco)
    gg = prep_dataset_coco(data_set=f_merge, ld_matrix=ld_abcg2,hwe_variance = F,exact=T, var_y=1.6421)
    gg2 = prep_dataset_coco(data_set=f_merge, ld_matrix=ld_abcg2,hwe_variance = F,exact=F, var_y=1.61078)
    get_ld(c("rs2231142","rs3114020","rs2622629","rs2054576"),gg)
    diag(ld_abcg2) = diag(ld_abcg2) + 0.5

    #f_merge$Freq1 = ifelse(f_merge$Freq1 > 0.5, 1 - f_merge$Freq1, f_merge$Freq1)
    #f_merge$FREQ1 = ifelse(f_merge$FREQ1 > 0.5, 1 - f_merge$FREQ1, f_merge$FREQ1)
    gg$exact = T
    #gg$neff = (gg$var_y - gg$var *gg$b^2)/(gg$var *gg$se^2) + 1



    lel = stepwise_coco(gg, joint=F,p_value_threshold =9.186111e-06)
    get_ld(c("rs2231142","rs3114020","rs2622629","rs2728126"),gg)