-
Notifications
You must be signed in to change notification settings - Fork 0
/
Section5_Genomic_Prediction.Rmd
512 lines (417 loc) · 15.2 KB
/
Section5_Genomic_Prediction.Rmd
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
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
---
title: "Section5 Genomic Prediction"
author: "Ye Bi"
date: "`r Sys.Date()`"
output: html_document
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
knitr::opts_knit$set(root.dir = "~/OneDrive - Virginia Tech/Research/Codes/research/RiceUNLMetabolites/Prediction/Prediction")
path.met = "../../Pheno"
path.trait = "../../Trait"
path.output = "./outputs"
path.geno = "../../Geno"
```
### Loading packages
```{r, message=F}
library(rrBLUP)
library(BGLR)
library(ggplot2)
library(ggpubr)
library(tidyverse)
```
# 1. loading data
## 1.1 Read met data
```{r, echo = T, eval = T}
met.con <- read.csv(file=file.path(path.met, "met.rr.con.csv")) #192*75
met.trt <- read.csv(file=file.path(path.met, "met.rr.trt.csv")) #188*75
```
## 1.2 Trait data: Major axis, Minor axis and Perimeter data
```{r, echo = TRUE, eval = T}
load(file = file.path(path.trait, "trait.con.Rdata"))
load(file = file.path(path.trait, "trait.trt.Rdata"))
trait.con$NSFTV_ID <- as.factor(trait.con$NSFTV_ID)
trait.trt$NSFTV_ID <- as.factor(trait.trt$NSFTV_ID)
```
## 1.3 Read SNP data
```{r, eval = F}
load(file=file.path(path.geno, "geno.rr.con.RData")) # 192 385118
load(file=file.path(path.geno, "geno.rr.trt.RData")) # 188 389854
```
# 2. Prediction for control
## 2.1 GBLUP and GMBLUP
```{r, echo=T, eval=F}
# Scale traits (Normalize)
y0 = scale(trait.con[,-c(1:2)], center = T, scale = T)
# G matrix
Gcs = scale(geno_con, center = T, scale = T)
Gchnt = tcrossprod(Gcs)/ncol(Gcs)
EVD_Gchnt <- eigen(Gchnt)
# M matrix
# M matrix
rownames(met.con) = met.con$NSFTV_ID
Mcs = scale(met.con[,-c(1:2)])
Mchnt = tcrossprod(Mcs) # for Bayesian MBLUP
Mchnt2 = Mchnt/mean(diag(Mchnt))
EVD_Mchnt <- eigen(Mchnt2)
#GBLUP
ETA1 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS')
)
#MBLUP
ETA2 <- list(
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
#GMBLUP
ETA3 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS'),
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
# set cross-validation parameters
ntst <- 38 # test 192*0.2
ntrn <- 192-38 # reference 192*0.8
nCV <- 100 # times of CV
traits <- c("MajorAxis", "MinorAxis", "Perimeter")
# predictive correlation
corR1 <- matrix(0, ncol = ncol(y0), nrow = nCV) #OLS
colnames(corR1) = traits
corR2 <- matrix(0, ncol = ncol(y0), nrow = nCV) #MBLUP
colnames(corR2) = traits
corR3 <- matrix(0, ncol = ncol(y0), nrow = nCV) #BayesC
colnames(corR3) = traits
# CV
for (i in 1:nCV) {
# i=1
# random-sampling to decide testing & reference accessions
set.seed(100 + i)
index <- sample(1:nrow(y0), size=ntst) # random sampling
y <- y0
y[index,] <- NA
for (j in 1:ncol(y0)){
cat("Now running nCV = ", i,"trait = ", j, "\n")
ySingle <- y[,j]
#GBLUP
fit1 <- BGLR(y=ySingle, ETA=ETA1, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F, saveAt = 'GBLUP_')
pred1 = fit1$yHat
corR1[i,j] <- cor(pred1[index], y0[index,j])
#GMBLUP
fit3 <- BGLR(y=ySingle, ETA=ETA3, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F, saveAt = 'GMBLUP_')
pred3 = fit3$yHat
corR3[i,j] <- cor(pred3[index], y0[index,j])
}}
save(corR1, corR3, file=file.path(path.output, "control_GBLUP_GMBLUP.rda"))
```
### 2.1.1 Calculate colmean
```{r}
load(file=file.path(path.output, "control_GBLUP_GMBLUP.rda"))
colnames(corR1) = c("Major_Axis", "Minor_Axis", "Perimeter")
round(colMeans(corR1), digits = 4) #GBLUP
# colnames(corR2) = c("Major_Axis", "Minor_Axis", "Perimeter")
# round(colMeans(corR2), digits = 4)#MBLUP
colnames(corR3) = c("Major_Axis", "Minor_Axis", "Perimeter")
round(colMeans(corR3), digits = 4) #GMBLUP
```
### 2.1.2 Heritability
```{r, echo=TRUE, eval = FALSE}
# Scale traits (Normalize)
y0 = scale(trait.con[,-c(1:2)], center = T, scale = T)
# G matrix
Gcs = scale(geno_con, center = T, scale = T)
Gchnt = tcrossprod(Gcs)/ncol(Gcs)
EVD_Gchnt <- eigen(Gchnt)
# M matrix
# M matrix
Mcs = scale(met.con[,-c(1:2)]) #since met.rr was scaled when I did BLUP correction, so here no necessary to double normalization.
Mchnt = tcrossprod(Mcs) # for Bayesian MBLUP
Mchnt2 = Mchnt/mean(diag(Mchnt))
EVD_Mchnt <- eigen(Mchnt2)
#GBLUP
ETA1 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS')
)
#MBLUP
ETA2 <- list(
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
#GMBLUP
ETA3 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS'),
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
for (j in 1:3){
cat("Now running trait: ", colnames(y0)[j], "\n")
set.seed(100+j)
ySingle <- y0[,j]
#GBLUP
fit1 <- BGLR(y=ySingle, ETA=ETA1, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit1$varE
varM = fit1$ETA$G$varU
h2M = varM/(varM+varE)
cat("fit1 heritability for genomics is: ", round(h2M, 4), "\n")
# MBLUP
fit2 <- BGLR(y=ySingle, ETA=ETA2, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit2$varE
varM = fit2$ETA$MET$varU
h2M = varM/(varM+varE)
cat("fit2 heritability for metabolites is: ", round(h2M,4), "\n")
#GMBLUP
fit3 <- BGLR(y=ySingle, ETA=ETA3, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit3$varE
varM = fit3$ETA$MET$varU
varG = fit3$ETA$G$varU
h2G = varG/(varG+varM+varE)
cat("fit3 heritability for genomics is: ", round(h2G,4), "\n")
h2M= varM/(varG+varE+varM)
cat("fit3 heritability for metabolites is: ", round(h2M,4), "\n")
}
```
Now running trait: MajorAxis
fit1 heritability for genomics is: 0.7836
fit2 heritability for metabolites is: 0.3254
fit3 heritability for genomics is: 0.6988
fit3 heritability for metabolites is: 0.0893
Now running trait: MinorAxis
fit1 heritability for genomics is: 0.7545
fit2 heritability for metabolites is: 0.4351
fit3 heritability for genomics is: 0.6312
fit3 heritability for metabolites is: 0.1336
Now running trait: Perimeter
fit1 heritability for genomics is: 0.7533
fit2 heritability for metabolites is: 0.3307
fit3 heritability for genomics is: 0.6257
fit3 heritability for metabolites is: 0.1167
# 3. Prediction for stress
## 3.1 GBLUP and GMBLUP
```{r, echo=T, eval=F}
# Scale traits (Normalize)
y0 = scale(trait.trt[,-c(1:2)], center = T, scale = T)
# G matrix
Gcs = scale(geno_trt, center = T, scale = T)
Gchnt = tcrossprod(Gcs)/ncol(Gcs)
EVD_Gchnt <- eigen(Gchnt)
# M matrix
# M matrix
Mcs = scale(met.trt[,-c(1:2)])
Mchnt = tcrossprod(Mcs) # for Bayesian MBLUP
Mchnt2 = Mchnt/mean(diag(Mchnt))
EVD_Mchnt <- eigen(Mchnt2)
#GBLUP
ETA1 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS')
)
#MBLUP
ETA2 <- list(
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
#GMBLUP
ETA3 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS'),
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
# set cross-validation parameters
ntst <- 38 # test 188*0.2
ntrn <- 188-38 # reference 188*0.8
nCV <- 100 # times of CV
traits <- c("MajorAxis", "MinorAxis", "Perimeter")
# predictive correlation
corR1 <- matrix(0, ncol = ncol(y0), nrow = nCV) #OLS
colnames(corR1) = traits
corR3 <- matrix(0, ncol = ncol(y0), nrow = nCV) #BayesC
colnames(corR3) = traits
# CV
for (i in 1:nCV) {
# random-sampling to decide testing & reference accessions
set.seed(100 + i)
index <- sample(1:nrow(y0), size=ntst) # random sampling
y <- y0
y[index,] <- NA
for (j in 1:ncol(y0)){
cat("Now running nCV = ", i,"trait = ", j, "\n")
ySingle <- y[,j]
#GBLUP
fit1 <- BGLR(y=ySingle, ETA=ETA1, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F, saveAt = 'GBLUP_')
pred1 = fit1$yHat
corR1[i,j] <- cor(pred1[index], y0[index,j])
#GMBLUP
fit3 <- BGLR(y=ySingle, ETA=ETA3, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F, saveAt = 'GMBLUP_')
pred3 = fit3$yHat
corR3[i,j] <- cor(pred3[index], y0[index,j])
}}
save(corR1, corR3, file=file.path(path.output, "stress_GBLUP_GMBLUP.rda"))
```
### 3.1.1 Calculate colmean
```{r}
load(file=file.path(path.output, "stress_GBLUP_GMBLUP.rda"))
colnames(corR1) = c("Major_Axis", "Minor_Axis", "Perimeter")
round(colMeans(corR1), digits = 4) #GBLUP
# colnames(corR2) = c("Major_Axis", "Minor_Axis", "Perimeter")
# round(colMeans(corR2), digits = 4)#MBLUP
colnames(corR3) = c("Major_Axis", "Minor_Axis", "Perimeter")
round(colMeans(corR3), digits = 4) #GMBLUP
```
### 3.1.2 Heritability
```{r, echo=TRUE, eval = FALSE}
# Scale traits (Normalize)
y0 = scale(trait.trt[,-c(1:2)], center = T, scale = T)
# G matrix
Gcs = scale(geno_trt, center = T, scale = T)
Gchnt = tcrossprod(Gcs)/ncol(Gcs)
EVD_Gchnt <- eigen(Gchnt)
# M matrix
# M matrix
Mcs = scale(met.trt[,-c(1:2)]) #since met.rr was scaled when I did BLUP correction, so here no necessary to double normalization.
Mchnt = tcrossprod(Mcs) # for Bayesian MBLUP
Mchnt2 = Mchnt/mean(diag(Mchnt))
EVD_Mchnt <- eigen(Mchnt2)
#GBLUP
ETA1 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS')
)
#MBLUP
ETA2 <- list(
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
#GMBLUP
ETA3 <- list(
G = list(V=EVD_Gchnt$vectors, d=EVD_Gchnt$values, model='RKHS'),
MET = list(V=EVD_Mchnt$vectors, d=EVD_Mchnt$values, model='RKHS')
)
for (j in 1:3){
cat("Now running trait: ", colnames(y0)[j], "\n")
set.seed(100+j)
ySingle <- y0[,j]
#GBLUP
fit1 <- BGLR(y=ySingle, ETA=ETA1, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit1$varE
varM = fit1$ETA$G$varU
h2M = varM/(varM+varE)
cat("fit1 heritability for genomics is: ", round(h2M, 4), "\n")
# MBLUP
fit2 <- BGLR(y=ySingle, ETA=ETA2, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit2$varE
varM = fit2$ETA$MET$varU
h2M = varM/(varM+varE)
cat("fit2 heritability for metabolites is: ", round(h2M,4), "\n")
#GMBLUP
fit3 <- BGLR(y=ySingle, ETA=ETA3, nIter = 30000,
burnIn = 10000, thin = 5, verbose = F)
varE = fit3$varE
varM = fit3$ETA$MET$varU
varG = fit3$ETA$G$varU
h2G = varG/(varG+varM+varE)
cat("fit3 heritability for genomics is: ", round(h2G,4), "\n")
h2M= varM/(varG+varE+varM)
cat("fit3 heritability for metabolites is: ", round(h2M,4), "\n")
}
```
# 4. Draw plots
```{r}
######################################
###loading MBLUP and GBLUP GMBLUP#####
######################################
load(file=file.path(path.output, "control_OLS_MBLUP_BayesC.rda"))
corR_MBLUP_con = corR2
load(file=file.path(path.output, "control_GBLUP_GMBLUP.rda"))
corR_GBLUP_con = corR1
corR_GMBLUP_con = corR3
load(file=file.path(path.output, "stress_OLS_MBLUP_BayesC.rda"))
corR_MBLUP_trt = corR2
load(file=file.path(path.output, "stress_GBLUP_GMBLUP.rda"))
corR_GBLUP_trt = corR1
corR_GMBLUP_trt = corR3
###############################################
## Percentage Compare GBLUP, GMBLUP, MBLUP#####
###############################################
GBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_GBLUP_con),cbind.data.frame(Treatment = "Stress", corR_GBLUP_trt))
MBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_MBLUP_con),cbind.data.frame(Treatment = "Stress", corR_MBLUP_trt))
GMBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_GMBLUP_con),cbind.data.frame(Treatment = "Stress", corR_GMBLUP_trt))
traits = c("MajorAxis", "MinorAxis", "Perimeter")
# labels = c(" Grain length", " Grain width", "Grain perimeter")
m=1
traitL <- list()
for( i in traits){
BLUPcorR <- data.frame(GBLUP[,c("Treatment",i)], MBLUP[, i], GMBLUP[,i])
BLUPcorR$Treatment <- as.factor(BLUPcorR$Treatment)
colnames(BLUPcorR) <- c("Treatment", "GBLUP", "MBLUP", "GMBLUP")
models <- c( "MBLUP", "GMBLUP", "GBLUP")
comb_mol <- matrix(c("GBLUP", "MBLUP", "MBLUP", "GMBLUP", "GBLUP", "GMBLUP"), 2, 3)
comb_mol_con <- rbind.data.frame(comb_mol, NA)
comb_mol_trt <- rbind.data.frame(comb_mol, NA) #first row - second rw.
BLUPcorR_con <- BLUPcorR %>% filter(Treatment == "Control")
BLUPcorR_trt <- BLUPcorR %>% filter(Treatment == "Stress")
for (j in 1:3){
# j=1
x=comb_mol[1,j]
y=comb_mol[2,j]
comb_mol_con[3,j] <- sum(BLUPcorR_con[,x]-BLUPcorR_con[,y]>0)
comb_mol_trt[3,j] <- sum(BLUPcorR_trt[,x]-BLUPcorR_trt[,y]>0)
}
traitL[[m]] <- list(control = comb_mol_con, stress = comb_mol_trt)
m = m+1
}
names(traitL) <- traits
GtraitL <- traitL
save(GtraitL, file="./compare_genomic_models_percentage.Rdata")
load("./compare_genomic_models_percentage.Rdata")
####################
## Draw plots#######
####################
GBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_GBLUP_con),cbind.data.frame(Treatment = "Stress", corR_GBLUP_trt))
MBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_MBLUP_con),cbind.data.frame(Treatment = "Stress", corR_MBLUP_trt))
GMBLUP <- rbind(cbind.data.frame(Treatment = "Control", corR_GMBLUP_con),cbind.data.frame(Treatment = "Stress", corR_GMBLUP_trt))
traits = c("MajorAxis", "MinorAxis", "Perimeter")
labels = c(" Grain length", " Grain width", "Grain perimeter")
pp <- list()
t=1
for( i in traits){
BLUPcorR <- data.frame(GBLUP[,c("Treatment",i)], MBLUP[, i], GMBLUP[,i])
BLUPcorR$Treatment <- as.factor(BLUPcorR$Treatment)
colnames(BLUPcorR) <- c("Treatment", "GBLUP", "MBLUP", "GMBLUP")
models <- c( "MBLUP", "GMBLUP", "GBLUP")
comb_mol <- matrix(c("GBLUP", "MBLUP", "MBLUP", "GMBLUP", "GBLUP", "GMBLUP"), 2, 3)
temp = GtraitL[[i]]
p=list()
aa=array()
for (j in 1:3){
BLUPcorR$Treatment <- rep(c(paste0("Control: ", temp$control[3, j],"%"), paste0("HNT: ", temp$stress[3, j], "%")), each =100)
p[[j]] <- ggplot(data = BLUPcorR, aes_string(comb_mol[1,j],comb_mol[2,j])) +
coord_fixed() +
scale_x_continuous(limits=c(-0.2, 0.9), breaks=c(-0.2,0,0.2,0.4,0.6,0.8)) +
scale_y_continuous(limits=c(-0.2, 0.9), breaks=c(-0.2,0,0.2,0.4,0.6,0.8)) +
geom_point(size=2, alpha=0.5, aes(colour=Treatment)) +
geom_abline(intercept = 0, slope = 1) + theme_bw() +
labs(x=comb_mol[1,j],y=comb_mol[2,j])+
theme(legend.position=c(0.815,0.08))+
# scale_color_discrete(name="") +
theme(legend.key.size = unit(0.01, "cm"), legend.title = element_blank())+
theme(legend.background = element_rect(fill="aliceblue"))+
theme(legend.text = element_text(size=10, face="bold"))
# print(p[[j]])
}
# plot_Ave <- ggarrange(p[[1]],p[[2]],p[[3]], labels=i, nrow = 1, ncol = 3, common.legend = F)
# print(plot_Ave)
pp[[t]] <- ggarrange(p[[1]],p[[2]],p[[3]], labels=labels[t], nrow = 1, ncol = 3, common.legend = F)
t=t+1
# dev.print(pdf, file = file.path(path.output, paste0(i,"_GBLUP_MBLUP_GMBLUP_comb.pdf")), height = 5, width = 12)
}
plot1 <- ggarrange(pp[[1]],pp[[2]],pp[[3]], nrow = 3, ncol = 1)
print(plot1)
dev.print(pdf, file = file.path(path.output, "GBLUP_MBLUP_GMBLUP_comb.pdf"), height = 14, width = 12)
```
## 5. Montel test for M and G matrix
```{r eval=F}
library(vegan)
G1 = dist(Gchnt)
M1 = dist(Mchnt)
mantel(M1, G1, method="pearson", permutations=999, strata = NULL,
na.rm = FALSE)
```