-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy path5.RKHS_SSI_Andean_predictionSet.R
235 lines (169 loc) · 5.83 KB
/
5.RKHS_SSI_Andean_predictionSet.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
223
224
225
226
227
228
229
230
231
232
233
234
235
#!/usr/bin/env Rscript
#SBATCH --time=00:20:00
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=16G
#SBATCH --constraint="[intel18|amd20|amd22]"
#SBATCH --array=1-300
#SBATCH -A data-machine
rm(list=ls())
library(SFSI)
library(BGLR)
library(tidyverse)
load('YBC_GWASAssistedGP.RData')
job <- as.integer(Sys.getenv("SLURM_ARRAY_TASK_ID", "1"))
JOBS <- expand.grid(rep=1:100, trait=1:3) # Fe, Zn, Yield 2019
trait <- as.vector(JOBS[job,"trait"])
rep <- as.vector(JOBS[job,"rep"])
rownames(YBC_GWASAssistedGP$pheno) <- YBC_GWASAssistedGP$pheno$taxa
traints_pre19 <- c("Yield_MI19", "Fe_MI19", "Zn_MI19")
index <- which(colnames(YBC_GWASAssistedGP$pheno) %in% traints_pre19)
Y = YBC_GWASAssistedGP$pheno[,index]
M = YBC_GWASAssistedGP$geno
# Andean info
index_andean <- which(YBC_GWASAssistedGP$pheno$Subpop_Sadohara_etal2022 =="Andean")
index_YA <- grep("Y1",rownames(M))
index <- c(index_andean,index_YA)
# Keep Andean and bi-parental crosses
Y = Y[index,]
M = M[index,]
# Remove markers with no variation
M = M[,apply(M,2,function(x) length(unique(x)))>1]
M <- scale(M)/sqrt(ncol(M))
G <- tcrossprod(M)
D <-as.matrix(dist(M,method="euclidean"))^2 #euclidian distance
D <-D/mean(D)
index <- which(!is.na(Y[,trait])) # remove NA
Y = Y[index,]
G = G[index,index]
D = D[index,index]
trait_name <- colnames(Y)[trait]
y <- Y[,trait]
n <- length(y)
index <- grep('YBC', rownames(Y))
# set training and test sets
trn0 = index
tst0 <- seq_along(1:n)[-trn0]
Y1608 = grep('Y1608', rownames(Y))
Y1608_10 = sample(Y1608, ceiling(length(Y1608)*0.1))
Y1608_20 = sample(Y1608, ceiling(length(Y1608)*0.2))
Y1608_30 = sample(Y1608, ceiling(length(Y1608)*0.3))
Y1609 = grep('Y1609', rownames(Y))
Y1609_10 = sample(Y1609, ceiling(length(Y1609)*0.1))
Y1609_20 = sample(Y1609, ceiling(length(Y1609)*0.2))
Y1609_30 = sample(Y1609, ceiling(length(Y1609)*0.3))
Y1612 = grep('Y1612', rownames(Y))
Y1612_10 = sample(Y1612, ceiling(length(Y1612)*0.1))
Y1612_20 = sample(Y1612, ceiling(length(Y1612)*0.2))
Y1612_30 = sample(Y1612, ceiling(length(Y1612)*0.3))
Y1701 = grep('Y1701', rownames(Y))
Y1701_10 = sample(Y1701, ceiling(length(Y1701)*0.1))
Y1701_20 = sample(Y1701, ceiling(length(Y1701)*0.2))
Y1701_30 = sample(Y1701, ceiling(length(Y1701)*0.3))
Y1702 = grep('Y1702', rownames(Y))
Y1702_10 = sample(Y1702, ceiling(length(Y1702)*0.1))
Y1702_20 = sample(Y1702, ceiling(length(Y1702)*0.2))
Y1702_30 = sample(Y1702, ceiling(length(Y1702)*0.3))
Y1703 = grep('Y1703', rownames(Y))
Y1703_10 = sample(Y1703, ceiling(length(Y1703)*0.1))
Y1703_20 = sample(Y1703, ceiling(length(Y1703)*0.2))
Y1703_30 = sample(Y1703, ceiling(length(Y1703)*0.3))
trn10 = c(trn0, Y1608_10,Y1609_10, Y1612_10,Y1701_10,
Y1702_10,Y1703_10)
trn20 = c(trn0, Y1608_20,Y1609_20, Y1612_20,Y1701_20,
Y1702_20,Y1703_20)
trn30 = c(trn0, Y1608_30,Y1609_30, Y1612_30,Y1701_30,
Y1702_30,Y1703_30)
tst10 <- seq_along(1:n)[-trn10]
tst20 <- seq_along(1:n)[-trn20]
tst30 <- seq_along(1:n)[-trn30]
# Calculate variance components ratio using training data
yNA = y
yNA[tst0] = NA
fm0 = fitBLUP(yNA,K=G)
GBLUP_0 = cor(fm0$u[tst0],y[tst0])
yNA = y
yNA[tst10] = NA
fm0 = fitBLUP(yNA,K=G)
GBLUP_10 = cor(fm0$u[tst0],y[tst0])
yNA = y
yNA[tst20] = NA
fm0 = fitBLUP(yNA,K=G)
GBLUP_20 = cor(fm0$u[tst0],y[tst0])
yNA = y
yNA[tst30] = NA
fm0 = fitBLUP(yNA,K=G)
GBLUP_30 = cor(fm0$u[tst0],y[tst0])
##### SSI
#0%
fm1 = SSI.CV(y,K=G,trn=trn0,
nCV=10,name="5 5CV", nfolds = 10)
lambda = summary(fm1)$optCOR["lambda"]
# Fit the index with the obtained lambda
fm2 = SSI(y,K=G,trn=trn0, tst=tst0,lambda=lambda)
SSI_0 = summary(fm2)$optCOR
#10%
fm1 = SSI.CV(y,K=G,trn=trn10,
nCV=10,name="5 5CV", nfolds = 10)
lambda = summary(fm1)$optCOR["lambda"]
# Fit the index with the obtained lambda
fm2 = SSI(y,K=G,trn=trn10, tst=tst10,lambda=lambda)
SSI_10 = summary(fm2)$optCOR
#20%
fm1 = SSI.CV(y,K=G,trn=trn20,
nCV=10,name="5 5CV", nfolds = 10)
lambda = summary(fm1)$optCOR["lambda"]
# Fit the index with the obtained lambda
fm2 = SSI(y,K=G,trn=trn20, tst=tst20,lambda=lambda)
SSI_20 = summary(fm2)$optCOR
#30%
fm1 = SSI.CV(y,K=G,trn=trn30,
nCV=10,name="5 5CV", nfolds = 10)
lambda = summary(fm1)$optCOR["lambda"]
# Fit the index with the obtained lambda
fm2 = SSI(y,K=G,trn=trn30, tst=tst30,lambda=lambda)
SSI_30 = summary(fm2)$optCOR
## KA
yNA = y
yNA[tst0] = NA
h<- c(.02,1,5) # bandwidth kernels
KList<-list()
for(i in 1:length(h)){ # Avering kernel
KList[[i]]<-list(K=exp(-h[i]*D),model='RKHS')
}
fmKA<-BGLR(y=yNA,ETA=KList,
nIter=12000,burnIn=2000,
saveAt=paste0("KA_preset0_",trait_name,"_rep_",rep))
KA_0 = cor(y[tst0],fmKA$yHat[tst0],
use = 'pairwise.complete.obs')
yNA = y
yNA[tst10] = NA
fmKA<-BGLR(y=yNA,ETA=KList,
nIter=12000,burnIn=2000,
saveAt=paste0("KA_preset10_",trait_name,"_rep_",rep))
KA_10 = cor(y[tst10],fmKA$yHat[tst10],
use = 'pairwise.complete.obs')
#== 20 %
yNA = y
yNA[tst20] = NA
fmKA<-BGLR(y=yNA,ETA=KList,
nIter=12000,burnIn=2000,
saveAt=paste0("KA_preset20_",trait_name,"_rep_",rep))
KA_20 = cor(y[tst20],fmKA$yHat[tst20],
use = 'pairwise.complete.obs')
# ======= 30%
yNA = y
yNA[tst30] = NA
fmKA<-BGLR(y=yNA,ETA=KList,
nIter=12000,burnIn=2000,
saveAt=paste0("KA_preset30_",trait_name,"_rep_",rep))
KA_30 = cor(y[tst30],fmKA$yHat[tst30],
use = 'pairwise.complete.obs')
results = c(KA_0,KA_10,KA_20,KA_30,
SSI_0,SSI_10,SSI_20,SSI_30)
names(results) = c('KA_0','KA_10','KA_20','KA_30',
'SSI_0', 'SSI0_MSE', 'SSI0_df','SSI0_lambda',
'SSI_10','SSI10_MSE', 'SSI10_df','SSI10_lambda',
'SSI_20','SSI20_MSE', 'SSI20_df','SSI20_lambda',
'SSI_30','SSI30_MSE', 'SSI30_df','SSI30_lambda')
save(results,
file=paste0("results_tst_Andean_",trait_name,"_rep_",rep,".RData"))