GWAS分析中一般线性模型GML中snp的回归系数r2的计算

 

001、plink计算

root@PC1:/home/test# ls
gwas_test.map  gwas_test.ped
root@PC1:/home/test# plink --file gwas_test --assoc 1> /dev/null
root@PC1:/home/test# ls
gwas_test.map  gwas_test.ped  plink.log  plink.qassoc
root@PC1:/home/test# head plink.qassoc     ## 倒数第三列为r2
 CHR        SNP         BP    NMISS       BETA         SE         R2        T            P
   1       snp1       2802      541     -8.911      8.344   0.002111   -1.068        0.286
   1       snp2       2823      541      7.754      10.04   0.001104    0.772       0.4405
   1       snp3       4512      541      9.264      9.814    0.00165   0.9439       0.3456
   1       snp4      16529      541     -18.49      10.81   0.005401   -1.711      0.08769
   1       snp5      16578      541      5.661       9.93  0.0006026   0.5701       0.5689
   1       snp6      16579      541     -5.577      5.657     0.0018  -0.9858       0.3247
   1       snp7      16635      541      2.985      8.717  0.0002176   0.3425       0.7321
   1       snp8      20879      541      9.053       7.29   0.002853    1.242       0.2148
   1       snp9      20908      541      9.278      6.695    0.00355    1.386       0.1664

 

 

002、R语言计算

root@PC1:/home/test# ls
gwas_test.map  gwas_test.ped
root@PC1:/home/test# plink --file gwas_test --recode A 1> /dev/null
root@PC1:/home/test# ls
gwas_test.map  gwas_test.ped  plink.log  plink.raw
root@PC1:/home/test# cut -d " " -f 1,3-5 --complement plink.raw > num.raw
root@PC1:/home/test# ls
gwas_test.map  gwas_test.ped  num.raw  plink.log  plink.raw

 

library(data.table)
dat <- fread("num.raw", data.table = F)
dat[1:3,1:8]

r2 <- vector()

for (i in 3:12 ){
  reg <- lm(PHENOTYPE ~  1 + dat[,i], data=dat)
  r2[i - 2] <- summary(reg)$r.squared
}
r2

 

参考：https://blog.csdn.net/yijiaobani/article/details/122093938?app_version=5.6.0&code=app_1562916241&csdn_share_tail=%7B%22type%22%3A%22blog%22%2C%22rType%22%3A%22article%22%2C%22rId%22%3A%22122093938%22%2C%22source%22%3A%22liujiaxin2019%22%7D&ctrtid=6XokY&uLinkId=usr1mkqgl919blen&utm_source=app