setwd(dir="D:/mendelian test randomization/test")  #设置你自己的工作路径

# -加载需要用到的包，没安装的需要提前安装一下
library(TwoSampleMR)
library(data.table)
library(tidyverse)
library(readxl)
library(writexl)
library(mr.raps)
library(ieugwasr)
library(dplyr)
library(mendelianRandomization)

# 暴露，把"1.txt"这个文件放在你设置的工作路径下
name <- read.table("1.txt")
exp_dat_ids <- name[,1]
exps <- name[,1]

# 【本地自己导入，环境保持一致】
# 结局
out<-fread("结局.txt",header = T)#读取本地结局（这里将文件名换成你自己本地结局的文件名）
out$trait <- ''  #改成你所用到的结局的表型的名称
outcomeid <- out
#
dir.create(path = "mendelian test")  #在你的工作路径下创建一个名为"mendelian test"的文件夹，后面的结果都会生成在这个文件夹里面

#######以下为循环代码，不需要进行更改######################
for (qaq in 1:length(exp_dat_ids)) {{ # 
  exp_dat_id <- exp_dat_ids[qaq]
  exp <- exps[qaq]
  
  exp_data <- extract_instruments(
    outcomes = exp_dat_id,p1=1e-05,clump =F )
  
  library(ieugwasr)     
  
  d4<- ld_clump(
    #dat = X1,
    clump_kb = 10000,
    clump_r2 = 0.01,    #连锁不平衡的标准
    pop = "EUR",
    dplyr::tibble(rsid=exp_data$SNP, pval=exp_data$pval.exposure, id=exp_data$id.exposure),
    #plink.exe的位置
    plink_bin = "D:/mendelian randomization/一些技巧/本地clump/plink_win64_20230116/plink.exe",
    #EUR.bed的位置
    bfile = "D:/mendelian randomization/血清代谢物教学课程/依赖文件/本地LD依赖文件及MAF文件/1kg.v3/1kg.v3/EUR"
  )
  
  exp_data<-subset(exp_data,SNP %in% d4$rsid) 
  
  ####phenoscanner###
  confounder_list <- lapply(seq(1, nrow(exp_data), 100), function(start_index) {
    end_index <- min(start_index + 99, nrow(exp_data))
    batch_data <- exp_data[start_index:end_index, ]
    
    MendelianRandomization::phenoscanner(
      snpquery = unique(batch_data$SNP),
      catalogue = "GWAS",
      pvalue = 5e-8,
      proxies = "None",
      r2 = 0.8,
      build = 37
    )
  })
  
  # 合并结果
  confounder <- bind_rows(lapply(confounder_list, `[[`, "results"))
  if (nrow(confounder) > 0) {
    a=confounder[,c("snp","trait","p")];a
    a$trait <- tolower(a$trait)
    
    confounders <- "gggggghhhhh" 
    ###填入你要研究的混杂，以|隔开
    
    pattern <- paste0("^(", paste(confounders, collapse = "|"), ")$")
    
    c <- a %>%
      filter(grepl(pattern, trait))
    
    if (nrow(c) > 0) {
      c$phenotype <- FileNames[qaq]
      
      exp_data <- exp_data[!(exp_data$SNP %in% c$snp), ]
      openxlsx::write.xlsx(c,
                           file = paste0("mendelian test/", exp, "-confounders.xlsx"),
                           row.names = FALSE)
    }} else {
      exp_data <- exp_data
    } 
  
  
  if(length(exp_data[,1])>0){
    outcome_dat<-merge(exp_data,outcomeid,by.x = "SNP",by.y = "SNP")
    if(length(outcome_dat[,1])>0){  
      write.csv(outcome_dat,file = "d.csv")
      out_data <- read_outcome_data(
        snps = exp_data$SNP, 
        filename = "d.csv",
        sep = ",")
      
      ###排除与结局强相关
      out_data <- subset(out_data,pval.outcome>5e-8)
      
      dat <- TwoSampleMR::harmonise_data(
        exposure_dat = exp_data,
        outcome_dat = out_data)
      
      ####回文的直接去除
      dat <-subset(dat,mr_keep==TRUE)
      
      
      #####新的计算R2的代码
      dat <- get_f(dat, F_value = 10)
      
      ##############MR-PRESSO######################    
      if (nrow(dat) > 3) {
        mr_presso_res <- mr_Presso(dat, num = 300)
        mr_presso_main <- mr_presso_pval(mr_presso_res)
        dat <- mr_presso_snp(mr_presso_res, mr_presso_main, dat, type = "data")
        dat <-subset(dat,mr_keep==TRUE)
        
        resMRPRESSO=mr_presso_res[["Main MR results"]]
        resMRPRESSO
        global_test_p <- mr_presso_res[["MR-PRESSO results"]][["Global Test"]][["Pvalue"]]
        se1=sqrt(((resMRPRESSO[1,3])^2)/qchisq(Pval_raw <- resMRPRESSO[1,6],1,lower.tail=F))
        se2=sqrt(((beta_cor <- resMRPRESSO[2,3])^2)/qchisq(Pval_cor <- resMRPRESSO[2,6],1,lower.tail=F))
        resMRPRESSO <- resMRPRESSO %>%
          dplyr::mutate(se = c(se1,se2))
        
        #生成globe test_P和outliers
        outliers <- dat$SNP[mr_presso_res[["MR-PRESSO results"]][["Distortion Test"]][["Outliers Indices"]]]
        outliers = as.data.frame(outliers)
        outliers <- paste(outliers$outliers, collapse = ", ")
        global_test_p = as.data.frame(global_test_p)
        resMRPRESSO
        TTT <- as.data.frame(resMRPRESSO)
        TTT
        openxlsx::write.xlsx(TTT, paste0("mendelian test/",exp,"-MR-PRESSO.xlsx"))
      } 
      
      res <-choose_MR(dat=dat)
      res_hete <-purrr::map(.x=seq_along(res),
                            .f=~res[[.x]][[1]])
      res_hete <-do.call(rbind,res_hete)
      res_hete
      res1 <- generate_odds_ratios(res[[1]][[2]])
      res1
      res1$estimate <- paste0(
        format(round(res1$or, 2), nsmall = 2), " (", 
        format(round(res1$or_lci95, 2), nsmall = 2), "-",
        format(round(res1$or_uci95, 2), nsmall = 2), ")")
      res1
      print(paste0(exp,"_SNP数_",res1$nsnp[1]))
      resdata <- dat
      openxlsx::write.xlsx(dat,file = paste0("mendelian test/",exp,"-dat.xlsx"), rowNames = FALSE)
      
      openxlsx::write.xlsx(res1,paste0("mendelian test/",exp,"-res.xlsx"))
      
      
      ######steiger检验######
      res_steiger <- steiger_test(dat) 
      
      
      # ##二分类结局power计算
      N <- dat$samplesize.outcome[1]
      alpha <- 0.05
      R2xz <- sum(dat$R2)
      K <- (dat$ncase.outcome[1] / dat$ncontrol.outcome[1])
      if (length(dat[, 1]) == 1) {
        OR <- res1 %>% filter(method == "Wald ratio") %>% pull(or)
      } else if (length(dat[, 1]) > 1) {
        OR <-
          res1 %>%filter(grepl("Inverse variance weighted", method)) %>%pull(or)
      }
      epower = NA
      power <- results_binary(N, alpha, R2xz, K, OR, epower)
      
      library(magrittr)
      # Main result 
      res3 <- res1#[1:5,]
      res3 <- res3[,-c(10:14)]
      res4 <- tidyr::pivot_wider(
        res3,names_from ="method",names_vary = "slowest",
        values_from = c("b","se","pval","estimate") )
      
      col_names <- colnames(res4)
      
      new_col_names <- gsub("\\(.*\\)", "", col_names)
      
      colnames(res4) <- new_col_names
      
      ##steiger
      res_steiger2 <- dplyr::select(res_steiger,
                                    correct_causal_direction,steiger_pval)
      #Power
      power2 <- tidyr::pivot_wider(
        power,names_from ="Parameter",names_vary = "slowest",
        values_from = c("Value","Description") )
      power2 <- power2[,1]
      
      
      # Merge
      res_ALL <- cbind(res4, res_steiger2,power2
      )
      res_ALL$F <- mean(dat$F,na.rm=TRUE)
      res_ALL$R2 <- sum(dat$R2)
      
      
      
      if (length(dat[, 1]) > 0 && length(dat[, 1]) <= 2) {
        write.csv(res_ALL, file = paste0("mendelian test/", exp, "1.csv"), row.names = FALSE)
      } else {
        # res_hete <- TwoSampleMR::mr_heterogeneity(dat)
        res_plei <- TwoSampleMR::mr_pleiotropy_test(dat)
        res_leaveone <- mr_leaveoneout(dat)
        
        p1 <- mr_scatter_plot(res[[1]][[2]], dat)
        pdf(paste0("mendelian test/", exp, "_scatter.pdf"))
        print(p1[[1]])
        dev.off()
        
        res_single <- mr_singlesnp(dat,all_method)
        p2 <- mr_forest_plot(res_single)
        pdf(paste0("mendelian test/", exp, "_forest.pdf"))
        print(p2[[1]])
        dev.off()
        
        p3 <- mr_funnel_plot(res_single)
        pdf(paste0("mendelian test/", exp, "_funnel.pdf"))
        print(p3[[1]])
        dev.off()
        
        res_loo <- mr_leaveoneout(dat)
        pdf(paste0("mendelian test/", exp, "_leave_one_out.pdf"))
        print(mr_leaveoneout_plot(res_loo))
        dev.off()
        
        
        
        res_hete2 <- tidyr::pivot_wider(
          res_hete, names_from = "method", names_vary = "slowest",
          values_from = c("Q", "Q_df", "Q_pval")
        ) %>% 
          dplyr::select(-id.exposure, -id.outcome, -outcome, -exposure)
        res_hete2 <- res_hete2[, 4:6]
        
        res_plei2 <- dplyr::select(res_plei, egger_intercept, se, pval)
        
        res_ALL <- cbind(res_ALL, res_hete2, res_plei2)
        
        if (nrow(dat) > 3) {
          
          res_ALL$outliers <- outliers
          res_ALL <- cbind(res_ALL, global_test_p)
        }
        write.csv(res_ALL, file = paste0("mendelian test/", exp, ".csv"), row.names = FALSE)
      }
    }
  }
}
}



#把导出的路径改成自己的工作路径  
fs=list.files("工作路径/mendelian test", pattern = "csv",full.names = TRUE) 
df = map_dfr(fs, read.csv)
df <- df %>%
  mutate(fdr_Inverse.variance.weighted. = p.adjust(pval_Inverse.variance.weighted., method = "fdr")) %>%
  relocate(fdr_Inverse.variance.weighted., .after = pval_Inverse.variance.weighted.)
write.csv(df,"mendelian testres.csv")