-
Notifications
You must be signed in to change notification settings - Fork 1
/
pl8-seed1-gpu-U-qu01-bu1.R
222 lines (193 loc) · 8.12 KB
/
pl8-seed1-gpu-U-qu01-bu1.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
rm(list = ls())
set.seed(1)
library(BWMR)
library(TwoSampleMR)
source("/home/jzhaoaz/sim-BWMR/gsmr_md.R")
library(ggplot2)
library(MASS) # Needed when sample from multivariate normal distribution
sim_null_4GM_select_sd <- function(P, N_1, N_2, Pi_11, Pi_10, Pi_01, Pi_00, beta, rho, noise, threshold, q_u, b_u){
# genotype data
# X_1: Individual-SNP matrix for exposure
X_1 <- matrix(1, N_1, P)
X_1f <- matrix(runif(N_1*P), N_1, P)
# X_2: Individual-SNP matrix for outcome
X_2 <- matrix(1, N_2, P)
X_2f <- matrix(runif(N_2*P), N_2, P)
# X_1: Individual-SNP matrix for exposure (MR)
X_3 <- matrix(1, N_1, P)
X_3f <- matrix(runif(N_1*P), N_1, P)
# f: minor allele frequencies of P SNPs
f <- runif(P, 0.05, 0.5)
f2_1 <- matrix(rep(f^2, each = N_1), N_1, P)
f2_2 <- matrix(rep(f^2, each = N_2), N_2, P)
f0_1 <- matrix(rep((1 - f)^2, each = N_1), N_1, P)
f0_2 <- matrix(rep((1 - f)^2, each = N_2), N_2, P)
# assign 2
X_1[X_1f < f2_1] <- 2
X_2[X_2f < f2_2] <- 2
X_2[X_3f < f2_1] <- 2
# assign 0
X_1[(1 - X_1f) < f0_1] <- 0
X_2[(1 - X_2f) < f0_2] <- 0
X_3[(1 - X_3f) < f0_1] <- 0
# Sigma: covariance matrix for (gamma, Gamma), under Pi_11
s1 <- 1 / (P*(Pi_11+Pi_10))
s2 <- 1 / (P*(Pi_11+Pi_01))
cov <- rho*sqrt(s1*s2)
Sigma <- matrix(c(s1, cov, cov, s2), 2, 2)
# W: SNP-(gamma, Gamma) matrix (ground truth)
W <- matrix(0, P, 2)
W[1:(P*Pi_11), ] <- mvrnorm(P*Pi_11, rep(0, 2), Sigma)
W[(P*Pi_11+1):(P*(Pi_11+Pi_10)), 1] <- rnorm(P*Pi_10, 0, sqrt(1/(P*(Pi_11+Pi_10))))
W[(P*(Pi_11+Pi_10)+1):(P*(Pi_11+Pi_10+Pi_01)), 2] <- rnorm(P*Pi_01, 0, sqrt(1/(P*(Pi_11+Pi_01))))
W[, 2] <- beta*W[, 1] + W[, 2]
# Exposure data
y0 <- X_1%*%W[, 1]
y <- y0 + noise*sqrt((1-q_u)*drop(var(y0)))*rnorm(N_1)
# Oucome data
z0 <- X_2%*%W[, 2]
# z <- z0 + noise*sqrt((1-q_u)*drop(var(z0)))*rnorm(N_2)
# Exposure data (MR)
y03 <- X_3%*%W[, 1]
y3 <- y03 + noise*sqrt((1-q_u)*drop(var(y03)))*rnorm(N_1)
# confounder U
u <- noise*sqrt(q_u*drop((var(y0)+var(y03))/2))*rnorm(N_1)
y <- y + u
y3 <- y3 + u
z <- z0 + b_u*u + noise*sqrt(drop(var(z0)) - drop(var(b_u*u)))*rnorm(N_1)
# b.exposure, se.exposure
b.exposure <- rep(0, P)
se.exposure <- rep(0, P)
pval.exposure <- rep(0, P)
for(j in 1:P){
fit_lm <- lm(y~., data = data.frame(y, X_1[, j]))
b.exposure[j] <- summary(fit_lm)$coefficients[2, 1]
se.exposure[j] <- summary(fit_lm)$coefficients[2, 2]
pval.exposure[j] <- summary(fit_lm)$coefficients[2, 4]
}
# b.outcome, se.outcome
b.outcome <- rep(0, P)
se.outcome <- rep(0, P)
pval.outcome <- rep(0, P)
for(j in 1:P){
fit_lm <- lm(z~., data = data.frame(z, X_2[, j]))
b.outcome[j] <- summary(fit_lm)$coefficients[2, 1]
se.outcome[j] <- summary(fit_lm)$coefficients[2, 2]
pval.outcome[j] <- summary(fit_lm)$coefficients[2, 4]
}
# b.exposure, se.exposure (MR)
b.exposure.mr <- rep(0, P)
se.exposure.mr <- rep(0, P)
pval.exposure.mr <- rep(0, P)
for(j in 1:P){
fit_lm <- lm(y3~., data = data.frame(y3, X_3[, j]))
b.exposure.mr[j] <- summary(fit_lm)$coefficients[2, 1]
se.exposure.mr[j] <- summary(fit_lm)$coefficients[2, 2]
pval.exposure.mr[j] <- summary(fit_lm)$coefficients[2, 4]
}
res_bias <- list()
res_unbias <- list()
for (j in 1:length(threshold)) {
indx <- which(pval.exposure < threshold[j])
res_bias[[j]] <- data.frame(b.exposure = b.exposure[indx], b.outcome = b.outcome[indx],
se.exposure = se.exposure[indx], se.outcome = se.outcome[indx])
res_unbias[[j]] <- data.frame(b.exposure = b.exposure.mr[indx], b.outcome = b.outcome[indx],
se.exposure = se.exposure.mr[indx], se.outcome = se.outcome[indx])
}
# res_all <- data.frame(b.exposure=b.exposure.mr, b.outcome=b.outcome,
# se.exposure=se.exposure.mr, se.outcome=se.outcome)
res <- list('unbias'=res_unbias, 'bias'=res_bias, 'index'=indx)
}
Rp <- 100
L <- 1
pval.bwmr.unbias <- matrix(nrow = L, ncol = Rp)
pval.raps.unbias <- matrix(nrow = L, ncol = Rp)
pval.egger.unbias <- matrix(nrow = L, ncol = Rp)
pval.gsmr.unbias <- matrix(nrow = L, ncol = Rp)
b.bwmr.unbias <- matrix(nrow = L, ncol = Rp)
b.raps.unbias <- matrix(nrow = L, ncol = Rp)
b.egger.unbias <- matrix(nrow = L, ncol = Rp)
b.gsmr.unbias <- matrix(nrow = L, ncol = Rp)
pval.bwmr.bias <- matrix(nrow = L, ncol = Rp)
pval.raps.bias <- matrix(nrow = L, ncol = Rp)
pval.egger.bias <- matrix(nrow = L, ncol = Rp)
pval.gsmr.bias <- matrix(nrow = L, ncol = Rp)
b.bwmr.bias <- matrix(nrow = L, ncol = Rp)
b.raps.bias <- matrix(nrow = L, ncol = Rp)
b.egger.bias <- matrix(nrow = L, ncol = Rp)
b.gsmr.bias <- matrix(nrow = L, ncol = Rp)
for (i in 1:Rp) {
message(i)
res <- sim_null_4GM_select_sd(P = 10000, N_1 = 5000, N_2 = 5000,
Pi_11 = 0.08, Pi_10 = 0.02, Pi_01 = 0.02, Pi_00 = 0.88,
beta = 0.0, rho = 0, noise = sqrt(1/1), threshold = c(1e-5),
q_u = 0.1, b_u = 1)
for (k in 1:1){ # k thresholds
## unbias
dat <- res$unbias[[k]]
N <- dim(dat)[1]
message("N=", N)
# BWMR
res_bwmr <- try(BWMR(gammahat = dat$b.exposure, Gammahat = dat$b.outcome,
sigmaX = dat$se.exposure, sigmaY = dat$se.outcome))
if (inherits(res_bwmr, "try-error")) {next}
b.bwmr.unbias[k, i] <- res_bwmr$beta
pval.bwmr.unbias[k, i] <- res_bwmr$P_value
# GSMR
res_gsmr <- try(gsmr(bzx = dat$b.exposure, bzx_se = dat$se.exposure,
bzy = dat$b.outcome, bzy_se = dat$se.outcome,
ldrho = diag(N),
heidi_outlier_flag = T,
heidi_outlier_thresh=0.01,
nsnps_thresh = 1))
if (inherits(res_gsmr, "try-error")) {next}
b.gsmr.unbias[k, i] <- drop(res_gsmr$bxy)
pval.gsmr.unbias[k, i] <- drop(res_gsmr$bxy_pval)
## Two sample MR
dat$id.exposure <- rep("uVXQCX", nrow(dat))
dat$id.outcome <- rep("uVXQCX", nrow(dat))
dat$mr_keep <- rep(TRUE, nrow(dat))
dat$exposure <- rep('exposure', nrow(dat))
dat$outcome <- rep('outcome', nrow(dat))
colnames(dat) <- c("beta.exposure", "beta.outcome", "se.exposure", "se.outcome", "id.exposure", "id.outcome", "mr_keep", "exposure", "outcome")
result <- try(mr(dat, method_list=c("mr_raps", "mr_egger_regression")))
if (inherits(result, "try-error")) {next}
b.raps.unbias[k, i] <- result$b[1]
pval.raps.unbias[k, i] <- result$pval[1]
b.egger.unbias[k, i] <- result$b[2]
pval.egger.unbias[k, i] <- result$pval[2]
## bias
dat <- res$bias[[k]]
N <- dim(dat)[1]
# BWMR
res_bwmr <- try(BWMR(gammahat = dat$b.exposure, Gammahat = dat$b.outcome,
sigmaX = dat$se.exposure, sigmaY = dat$se.outcome))
if (inherits(res_bwmr, "try-error")) {next}
b.bwmr.bias[k, i] <- res_bwmr$beta
pval.bwmr.bias[k, i] <- res_bwmr$P_value
# GSMR
res_gsmr <- try(gsmr(bzx = dat$b.exposure, bzx_se = dat$se.exposure,
bzy = dat$b.outcome, bzy_se = dat$se.outcome,
ldrho = diag(N),
heidi_outlier_flag = T,
heidi_outlier_thresh=0.01,
nsnps_thresh = 1))
if (inherits(res_gsmr, "try-error")) {next}
b.gsmr.bias[k, i] <- drop(res_gsmr$bxy)
pval.gsmr.bias[k, i] <- drop(res_gsmr$bxy_pval)
## Two sample MR
dat$id.exposure <- rep("uVXQCX", nrow(dat))
dat$id.outcome <- rep("uVXQCX", nrow(dat))
dat$mr_keep <- rep(TRUE, nrow(dat))
dat$exposure <- rep('exposure', nrow(dat))
dat$outcome <- rep('outcome', nrow(dat))
colnames(dat) <- c("beta.exposure", "beta.outcome", "se.exposure", "se.outcome", "id.exposure", "id.outcome", "mr_keep", "exposure", "outcome")
result <- try(mr(dat, method_list=c("mr_raps", "mr_egger_regression")))
if (inherits(result, "try-error")) {next}
b.raps.bias[k, i] <- result$b[1]
pval.raps.bias[k, i] <- result$pval[1]
b.egger.bias[k, i] <- result$b[2]
pval.egger.bias[k, i] <- result$pval[2]
}
}
save.image("pl8-seed1-gpu-U-qu01-bu1.RData")