-
Notifications
You must be signed in to change notification settings - Fork 2
Expand file tree
/
Copy pathBayesian_models.R
More file actions
281 lines (218 loc) · 11.1 KB
/
Bayesian_models.R
File metadata and controls
281 lines (218 loc) · 11.1 KB
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
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
###############################################################################
# Article: Genomic selection with allele dosage in Panicum maximum (Jacq.) #
# Lara et al., 2018 #
# Submitted to G3 (Genes|Genomes|Genetics) #
###############################################################################
rm(list=ls(all=T))
gc()
library(foreach)
library(doMC)
registerDoMC(32) # be aware with this command line -> it is using 32 cores
library(BGLR)
library(dplyr)
library(coda)
library(cvTools)
library(AGHmatrix)
# Phenotipic data
means <- read.table(file="./blup_means.txt", sep = ",", dec = ".", header = TRUE)
head(means)
y.trait <- means$MOF
# Genotipic data
Z <- diag(nrow(means))
dim(Z)
load("./molecular_data.RData")
dim(X4) # 530 offsprings and 41424 markers
nIter <- 20000
burnIn <- 2000
##Cross-validation sets:####
ETA.BRR <- list(X=list(X=X4, model='BRR')) # BRR -> Gaussian Prior
ETA.BA <- list(X=list(X=X4, model='BayesA')) # Bayes A -> Scaled-t Prior
ETA.BB <- list(X=list(X=X4, model='BayesB')) # Bayes B -> Point of mass at zero + scaled-t slab
ETA.BC <- list(X=list(X=X4, model='BayesC')) # Bayes C -> Point of mass at zero + Gaussian slab
ETA.BL <- list(X=list(X=X4, model='BL')) # BLASSO -> Laplace prior
n_id = length(means$Offspring)
n_model = 5
s = 100
y_pred_list = NULL
y_pred_list = c(y_pred_list, 1)
y_pred.BRR = matrix(NA,n_id,s)
y_pred.BA = matrix(NA,n_id,s)
y_pred.BB = matrix(NA,n_id,s)
y_pred.BC = matrix(NA,n_id,s)
y_pred.BL = matrix(NA,n_id,s)
list_for = foreach(k=1:s, .export="y_pred_list") %dopar% {
## Sets: ####
seed_value = as.integer(paste("2001",k,sep="")) # Fix the seed to each iterations
set.seed(seed_value)
cv = cvFolds(n_id, K = 5) # Obtaining cross-validation subsets
cv_vals = as.matrix(unlist(cv["subsets"][1])) # Indexing subsets
ID_names = means$Offspring
out_set1 = cv_vals[unlist(cv['which'])==1,1] # Indexing subset1
out_set2 = cv_vals[unlist(cv['which'])==2,1] # Indexing subset2
out_set3 = cv_vals[unlist(cv['which'])==3,1] # Indexing subset3
out_set4 = cv_vals[unlist(cv['which'])==4,1] # Indexing subset4
out_set5 = cv_vals[unlist(cv['which'])==5,1] # Indexing subset5
out_sets = list(out_set1, out_set2, out_set3, out_set4, out_set5)
set1 = c(out_set2,out_set3,out_set4,out_set5) # out_set1 out
set1 = set1[order(set1,decreasing=F)]
set2 = c(out_set1,out_set3,out_set4,out_set5) # out_set2 out
set2 = set2[order(set2,decreasing=F)]
set3 = c(out_set1,out_set2,out_set4,out_set5) # out_set3 out
set3 = set3[order(set3,decreasing=F)]
set4 = c(out_set1,out_set2,out_set3,out_set5) # out_set4 out
set4 = set4[order(set4,decreasing=F)]
set5 = c(out_set1,out_set2,out_set3,out_set4) # out_set5 out
set5 = set5[order(set5,decreasing=F)]
# Evaluating with there is some ID common across out_set
all(!c(out_set1%in%out_set2,out_set1%in%out_set3,out_set1%in%out_set4,out_set1%in%out_set5,out_set2%in%out_set3,out_set2%in%out_set4,out_set2%in%out_set5,out_set3%in%out_set4,out_set3%in%out_set5,out_set4%in%out_set5)) #OK!
sets = list(set1=set1,set2=set2,set3=set3,set4=set4,set5=set5)
out_sets = list(out_set1=out_set1,out_set2=out_set2,out_set3=out_set3,out_set4=out_set4,out_set5=out_set5)
rm(out_set1,out_set2,out_set3,out_set4,out_set5,set1,set2,set3,set4,set5)
## BRR: ####
y_obs=matrix(NA,n_id,1)
y_obs[sets$set1]=y.trait[sets$set1]
ans.brr1 <- BGLR(y=y_obs, ETA=ETA.BRR, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BRR[out_sets$out_set1,k] <- ans.brr1$yHat[out_sets$out_set1]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set2]=y.trait[sets$set2]
ans.brr2 <- BGLR(y=y_obs, ETA=ETA.BRR, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BRR[out_sets$out_set2,k] <- ans.brr2$yHat[out_sets$out_set2]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set3]=y.trait[sets$set3]
ans.brr3 <- BGLR(y=y_obs, ETA=ETA.BRR, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BRR[out_sets$out_set3,k] <- ans.brr3$yHat[out_sets$out_set3]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set4]=y.trait[sets$set4]
ans.brr4 <- BGLR(y=y_obs, ETA=ETA.BRR, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BRR[out_sets$out_set4,k] <- ans.brr4$yHat[out_sets$out_set4]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set5]=y.trait[sets$set5]
ans.brr5 <- BGLR(y=y_obs, ETA=ETA.BRR, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BRR[out_sets$out_set5,k] <- ans.brr5$yHat[out_sets$out_set5]
y_pred_list = c(y_pred_list,c(y_pred.BRR[,k]))
## Bayes A: ####
y_obs=matrix(NA,n_id,1)
y_obs[sets$set1]=y.trait[sets$set1]
ans.ba1 <- BGLR(y=y_obs, ETA=ETA.BA, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BA[out_sets$out_set1,k] <- ans.ba1$yHat[out_sets$out_set1]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set2]=y.trait[sets$set2]
ans.ba2 <- BGLR(y=y_obs, ETA=ETA.BA, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BA[out_sets$out_set2,k] <- ans.ba2$yHat[out_sets$out_set2]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set3]=y.trait[sets$set3]
ans.ba3 <- BGLR(y=y_obs, ETA=ETA.BA, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BA[out_sets$out_set3,k] <- ans.ba3$yHat[out_sets$out_set3]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set4]=y.trait[sets$set4]
ans.ba4 <- BGLR(y=y_obs, ETA=ETA.BA, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BA[out_sets$out_set4,k] <- ans.ba4$yHat[out_sets$out_set4]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set5]=y.trait[sets$set5]
ans.ba5 <- BGLR(y=y_obs, ETA=ETA.BA, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BA[out_sets$out_set5,k] <- ans.ba5$yHat[out_sets$out_set5]
y_pred_list = c(y_pred_list,c(y_pred.BA[,k]))
## Bayes B: ####
y_obs=matrix(NA,n_id,1)
y_obs[sets$set1]=y.trait[sets$set1]
ans.bb1 <- BGLR(y=y_obs, ETA=ETA.BB, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BB[out_sets$out_set1,k] <- ans.bb1$yHat[out_sets$out_set1]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set2]=y.trait[sets$set2]
ans.bb2 <- BGLR(y=y_obs, ETA=ETA.BB, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BB[out_sets$out_set2,k] <- ans.bb2$yHat[out_sets$out_set2]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set3]=y.trait[sets$set3]
ans.bb3 <- BGLR(y=y_obs, ETA=ETA.BB, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BB[out_sets$out_set3,k] <- ans.bb3$yHat[out_sets$out_set3]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set4]=y.trait[sets$set4]
ans.bb4 <- BGLR(y=y_obs, ETA=ETA.BB, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BB[out_sets$out_set4,k] <- ans.bb4$yHat[out_sets$out_set4]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set5]=y.trait[sets$set5]
ans.bb5 <- BGLR(y=y_obs, ETA=ETA.BB, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BB[out_sets$out_set5,k] <- ans.bb5$yHat[out_sets$out_set5]
y_pred_list = c(y_pred_list,c(y_pred.BB[,k]))
## Bayes C: ####
y_obs=matrix(NA,n_id,1)
y_obs[sets$set1]=y.trait[sets$set1]
ans.bc1 <- BGLR(y=y_obs, ETA=ETA.BC, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BC[out_sets$out_set1,k] <- ans.bc1$yHat[out_sets$out_set1]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set2]=y.trait[sets$set2]
ans.bc2 <- BGLR(y=y_obs, ETA=ETA.BC, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BC[out_sets$out_set2,k] <- ans.bc2$yHat[out_sets$out_set2]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set3]=y.trait[sets$set3]
ans.bc3 <- BGLR(y=y_obs, ETA=ETA.BC, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BC[out_sets$out_set3,k] <- ans.bc3$yHat[out_sets$out_set3]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set4]=y.trait[sets$set4]
ans.bc4 <- BGLR(y=y_obs, ETA=ETA.BC, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BC[out_sets$out_set4,k] <- ans.bc4$yHat[out_sets$out_set4]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set5]=y.trait[sets$set5]
ans.bc5 <- BGLR(y=y_obs, ETA=ETA.BC, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BC[out_sets$out_set5,k] <- ans.bc5$yHat[out_sets$out_set5]
y_pred_list = c(y_pred_list,c(y_pred.BC[,k]))
## BLASSO: ####
y_obs=matrix(NA,n_id,1)
y_obs[sets$set1]=y.trait[sets$set1]
ans.bl1 <- BGLR(y=y_obs, ETA=ETA.BL, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BL[out_sets$out_set1,k] <- ans.bl1$yHat[out_sets$out_set1]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set2]=y.trait[sets$set2]
ans.bl2 <- BGLR(y=y_obs, ETA=ETA.BL, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BL[out_sets$out_set2,k] <- ans.bl2$yHat[out_sets$out_set2]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set3]=y.trait[sets$set3]
ans.bl3 <- BGLR(y=y_obs, ETA=ETA.BL, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BL[out_sets$out_set3,k] <- ans.bl3$yHat[out_sets$out_set3]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set4]=y.trait[sets$set4]
ans.bl4 <- BGLR(y=y_obs, ETA=ETA.BL, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BL[out_sets$out_set4,k] <- ans.bl4$yHat[out_sets$out_set4]
y_obs=matrix(NA,n_id,1)
y_obs[sets$set5]=y.trait[sets$set5]
ans.bl5 <- BGLR(y=y_obs, ETA=ETA.BL, nIter=nIter, burnIn=burnIn, thin=5, verbose = F)
y_pred.BL[out_sets$out_set5,k] <- ans.bl5$yHat[out_sets$out_set5]
y_pred_list = c(y_pred_list,c(y_pred.BL[,k]))
}
all(is.na(y_pred_list))
y_pred_matrix = matrix(NA,(n_id*n_model),s)
names = paste("G", rep(1:n_id,times=n_model),"_M",rep(1:n_model,each=n_id),sep="")
rownames(y_pred_matrix) <- names
colnames(y_pred_matrix) <- paste("s",1:s,sep="")
count = 1
for (i in 1:s) {
y_pred_matrix[,i] = list_for[[i]][-1]
}
y_pred.BRR_final = y_pred_matrix[grepl(c("M1"),rownames(y_pred_matrix)),]
y_pred.BA_final = y_pred_matrix[grepl(c("M2"),rownames(y_pred_matrix)),]
y_pred.BB_final = y_pred_matrix[grepl(c("M3"),rownames(y_pred_matrix)),]
y_pred.BC_final = y_pred_matrix[grepl(c("M4"),rownames(y_pred_matrix)),]
y_pred.BL_final = y_pred_matrix[grepl(c("M5"),rownames(y_pred_matrix)),]
for (i in 1:s) {
if(i==1) {
accuracy_intervals = matrix(NA,s,n_model)
press = matrix(NA,s,n_model)
}
accuracy_intervals[i,1] = cor(y_pred.BRR_final[,i], y.trait)
accuracy_intervals[i,2] = cor(y_pred.BA_final[,i], y.trait)
accuracy_intervals[i,3] = cor(y_pred.BB_final[,i], y.trait)
accuracy_intervals[i,4] = cor(y_pred.BC_final[,i], y.trait)
accuracy_intervals[i,5] = cor(y_pred.BL_final[,i], y.trait)
press[i,1] = sum((y.trait - y_pred.BRR_final[,i])^2)
press[i,2] = sum((y.trait - y_pred.BA_final[,i])^2)
press[i,3] = sum((y.trait - y_pred.BB_final[,i])^2)
press[i,4] = sum((y.trait - y_pred.BC_final[,i])^2)
press[i,5] = sum((y.trait - y_pred.BL_final[,i])^2)
}
dim(accuracy_intervals)
colnames(accuracy_intervals) <- c("BRR","BayesA", "BayesB", "BayesC", "BLASSO")
accuracy_intervals_mcmc = as.mcmc(accuracy_intervals)
HPDinterval(accuracy_intervals_mcmc, prob=0.95)
colMeans(accuracy_intervals)
colnames(press) <- c("BRR","BayesA", "BayesB", "BayesC", "BLASSO")
colMeans(press)