-
Notifications
You must be signed in to change notification settings - Fork 0
/
05-GetGEBVs.Rmd
271 lines (245 loc) · 11.3 KB
/
05-GetGEBVs.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
---
title: "Predict GEBV for NRCRI C2"
author: "wolfemd"
date: "2020-April-27"
output: workflowr::wflow_html
editor_options:
chunk_output_type: inline
---
```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE, message = FALSE, warning = F, tidy = T)
```
# Previous step
4. [Check prediction accuracy](04-CrossValidation.html): Evaluate prediction accuracy with cross-validation.
# Objective
**Current Step**
5. [Genomic prediction of GS C2](05-GetGBLUPs.html): Predict _genomic_ BLUPs (GEBV and GETGV) for all selection candidates using all available data.
```{bash, eval=F}
# activate multithread OpenBLAS
export OMP_NUM_THREADS=48
#export OMP_NUM_THREADS=88
#export OMP_NUM_THREADS=88
```
# Set-up training-testing data
```{r, eval=F}
library(tidyverse); library(magrittr);
A<-readRDS(file=here::here("output","Kinship_A_NRCRI_2020April27.rds"))
D<-readRDS(file=here::here("output","Kinship_A_NRCRI_2020April27.rds"))
#AA<-readRDS(file=here::here("output","Kinship_AA_NRCRI_2020April27.rds"))
AD<-readRDS(file=here::here("output","Kinship_AD_NRCRI_2020April27.rds"))
#DD<-readRDS(file=here::here("output","Kinship_DD_NRCRI_2020April27.rds"))
blups_iita<-readRDS(file=here::here("data","iita_blupsForCrossVal_outliersRemoved_73019.rds"))
blups_iita<-blups_iita %>%
dplyr::select(Trait,blups) %>%
unnest(blups) %>%
dplyr::select(-`std error`) %>%
filter(GID %in% rownames(A),
!grepl("TMS13F|TMS14F|TMS15F|2013_",GID)) # don't want IITA GS progenies
blups_nrcri<-readRDS(file=here::here("output","nrcri_blupsForModelTraining_2020April27.rds"))
blups_nrcri<-blups_nrcri %>%
dplyr::select(Trait,modelOutput) %>%
unnest(modelOutput) %>%
dplyr::select(Trait,BLUPs) %>%
unnest(BLUPs) %>%
filter(GID %in% rownames(A))
blups<-blups_nrcri %>%
nest(TrainingData=-Trait) %>%
mutate(Dataset="NRCRIalone") %>%
bind_rows(blups_nrcri %>%
bind_rows(blups_iita %>% filter(Trait %in% blups_nrcri$Trait)) %>%
nest(TrainingData=-Trait) %>%
mutate(Dataset="IITAaugmented"))
blups
```
# Prediction
## runGenomicPredictions function
```{r runGenomicPredictions, eval=F}
#' @param blups nested data.frame with list-column "TrainingData" containing BLUPs
#' @param modelType string, A, AD or ADE representing model with Additive-only, Add. plus Dominance, and Add. plus Dom. plus. Epistasis (AA+AD+DD), respectively.
#' @param grms list of GRMs. Any genotypes in the GRMs get predicted with, or without phenotypes. Each element is named either A, D, AA, AD, DD. Matrices supplied must match required by A, AD and ADE models. For ADE grms=list(A=A,D=D,AA=AA,AD=AD,DD=DD).
runGenomicPredictions<-function(blups,modelType,grms,ncores=1,gid="GID",...){
require(sommer);
runOnePred<-possibly(function(trainingdata,modelType,grms){
trainingdata[[paste0(gid,"a")]]<-factor(trainingdata[[gid]],levels=rownames(grms[["A"]]))
if(modelType %in% c("AD","ADE")){ trainingdata[[paste0(gid,"d")]]<-factor(trainingdata[[gid]],levels=rownames(grms[["D"]]))
if(modelType=="ADE"){
#trainingdata[[paste0(gid,"aa")]]<-factor(trainingdata[[gid]],levels=rownames(grms[["AA"]]))
trainingdata[[paste0(gid,"ad")]]<-factor(trainingdata[[gid]],levels=rownames(grms[["AD"]]))
diag(grms[["AD"]])<-diag(grms[["AD"]])+1e-06
#trainingdata[[paste0(gid,"dd")]]<-factor(trainingdata[[gid]],levels=rownames(grms[["DD"]]))
}
}
# Set-up random model statements
randFormula<-paste0("~vs(",gid,"a,Gu=A)")
if(modelType %in% c("AD","ADE")){
randFormula<-paste0(randFormula,"+vs(",gid,"d,Gu=D)")
if(modelType=="ADE"){
randFormula<-paste0(randFormula,"+vs(",gid,"ad,Gu=AD)")
# "+vs(",gid,"aa,Gu=AA)",
# "+vs(",gid,"dd,Gu=DD)")
}
}
# Fit genomic prediction model
fit <- mmer(fixed = drgBLUP ~1,
random = as.formula(randFormula),
weights = WT,
data=trainingdata)
# Gather the BLUPs
gblups<-tibble(GID=as.character(names(fit$U[[paste0("u:",gid,"a")]]$drgBLUP)),
GEBV=as.numeric(fit$U[[paste0("u:",gid,"a")]]$drgBLUP))
if(modelType %in% c("AD","ADE")){
gblups %<>% mutate(GEDD=as.numeric(fit$U[[paste0("u:",gid,"d")]]$drgBLUP))
if(modelType=="ADE"){
gblups %<>% mutate(#GEEDaa=as.numeric(fit$U[[paste0("u:",gid,"aa")]]$drgBLUP),
GEEDad=as.numeric(fit$U[[paste0("u:",gid,"ad")]]$drgBLUP))
#GEEDdd=as.numeric(fit$U[[paste0("u:",gid,"dd")]]$drgBLUP))
}
}
# Calc GETGVs
## Note that for modelType=="A", GEBV==GETGV
gblups %<>%
mutate(GETGV=rowSums(.[,grepl("GE",colnames(.))]))
varcomps<-summary(fit)$varcomp
out<-tibble(gblups=list(gblups),varcomps=list(varcomps))
return(out)
},otherwise = NA)
## Run models across all train-test splits
## Parallelize
require(furrr); plan(multiprocess); options(mc.cores=ncores);
predictions<-blups %>%
mutate(genomicPredOut=future_map(TrainingData,~runOnePred(trainingdata=.,
modelType=modelType,grms=grms)))
return(predictions)
}
```
cbsulm18 (88 cores; 512GB)
Model A
```{r, eval=F}
options(future.globals.maxSize= 1500*1024^2)
predModelA<-runGenomicPredictions(blups,modelType="A",grms=list(A=A),gid="GID",ncores=1)
saveRDS(predModelA,file = here::here("output","genomicPredictions_ModelA_NRCRI_2020April27.rds"))
```
Model ADE
```{r, eval=F}
options(future.globals.maxSize= 3000*1024^2)
predModelADE<-runGenomicPredictions(blups,modelType="ADE",grms=list(A=A,D=D,AD=AD),gid="GID",ncores=20)
saveRDS(predModelADE,file = here::here("output","genomicPredictions_ModelADE_NRCRI_2020April27.rds"))
```
## Plot Predictions
```{r}
library(tidyverse); library(magrittr);
predModelA<-readRDS(file = here::here("output","genomicPredictions_ModelA_NRCRI_2020April27.rds"))
predModelADE<-readRDS(file = here::here("output","genomicPredictions_ModelADE_NRCRI_2020April27.rds"))
```
```{r}
predModelA %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(-GETGV) %>% rename(GEBV_A=GEBV) %>%
left_join(predModelADE %>%
mutate(islgl=map_lgl(genomicPredOut,is.logical)) %>%
filter(islgl==FALSE) %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(Trait,Dataset,GID,GEBV) %>% rename(GEBV_ADE=GEBV)) %>%
mutate(GeneticGroup=NA,
GeneticGroup=ifelse(grepl("NR17F",GID,ignore.case = T),"C2a",
ifelse(grepl("C2bF",GID,ignore.case = T),"C2b",
ifelse(grepl("^F",GID) & !grepl("C1b",GID),"C1a",
ifelse(grepl("C1b",GID,
ignore.case = T),"C1b","TrainingPop"))))) %>%
ggplot(.,aes(x=GEBV_A,y=GEBV_ADE,color=GeneticGroup)) +
geom_point() + theme_bw() + geom_abline(slope=1, color='darkred') +
facet_wrap(~Trait, scales = 'free') +
labs(title="Compare GEBV from A-only model to GEBV from A+D+E model")
```
```{r}
predModelADE %>%
mutate(islgl=map_lgl(genomicPredOut,is.logical)) %>%
filter(islgl==FALSE) %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(Trait,Dataset,GID,GEBV,GETGV) %>%
mutate(GeneticGroup=NA,
GeneticGroup=ifelse(grepl("NR17F",GID,ignore.case = T),"C2a",
ifelse(grepl("C2bF",GID,ignore.case = T),"C2b",
ifelse(grepl("^F",GID) & !grepl("C1b",GID),"C1a",
ifelse(grepl("C1b",GID,
ignore.case = T),"C1b","TrainingPop"))))) %>%
ggplot(.,aes(x=GEBV,y=GETGV,color=GeneticGroup)) +
geom_point() + theme_bw() + geom_abline(slope=1, color='darkred') +
facet_wrap(~Trait, scales = 'free') +
labs(title="Model: ADE - Compare GEBV to GETGV")
```
## Write GEBVs
```{r predModelA}
predModelA %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(-GETGV) %>%
spread(Trait,GEBV) %>%
mutate(GeneticGroup=NA,
GeneticGroup=ifelse(grepl("NR17F",GID,ignore.case = T),"C2a",
ifelse(grepl("C2bF",GID,ignore.case = T),"C2b",
ifelse(grepl("^F",GID) & !grepl("C1b",GID),"C1a",
ifelse(grepl("C1b",GID,
ignore.case = T),"C1b","TrainingPop"))))) %>%
nest(GEBVs=-Dataset) %>%
mutate(write=map2(Dataset,GEBVs,~write.csv(x = .y, file = here::here("output",paste0("GEBV_NRCRI_",.x,"_ModelA_2020April27.rds")))))
```
```{r predModelADE}
## Format and write GEBV
predModelADE %>%
mutate(islgl=map_lgl(genomicPredOut,is.logical)) %>%
filter(islgl==FALSE) %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(Trait,Dataset,GID,GEBV) %>%
spread(Trait,GEBV) %>%
mutate(GeneticGroup=NA,
GeneticGroup=ifelse(grepl("NR17F",GID,ignore.case = T),"C2a",
ifelse(grepl("C2bF",GID,ignore.case = T),"C2b",
ifelse(grepl("^F",GID) & !grepl("C1b",GID),"C1a",
ifelse(grepl("C1b",GID,
ignore.case = T),"C1b","TrainingPop"))))) %>%
nest(GEBVs=-Dataset) %>%
mutate(write=map2(Dataset,GEBVs,~write.csv(x = .y, file = here::here("output",paste0("GEBV_NRCRI_",.x,"_ModelADE_2020April27.rds")))))
## Format and write GETGV
predModelADE %>%
mutate(islgl=map_lgl(genomicPredOut,is.logical)) %>%
filter(islgl==FALSE) %>%
dplyr::select(Trait,Dataset,genomicPredOut) %>%
unnest(genomicPredOut) %>%
select(-varcomps) %>%
unnest(gblups) %>%
select(Trait,Dataset,GID,GETGV) %>%
spread(Trait,GETGV) %>%
mutate(GeneticGroup=NA,
GeneticGroup=ifelse(grepl("NR17F",GID,ignore.case = T),"C2a",
ifelse(grepl("C2bF",GID,ignore.case = T),"C2b",
ifelse(grepl("^F",GID) & !grepl("C1b",GID),"C1a",
ifelse(grepl("C1b",GID,
ignore.case = T),"C1b","TrainingPop"))))) %>%
nest(GETGVs=-Dataset) %>%
mutate(write=map2(Dataset,GETGVs,
~write.csv(x = .y,
file = here::here("output",paste0("GETGV_NRCRI_",.x,"_ModelADE_2020April27.rds")))))
```
# Next step
6. [Estimate genetic gain](06-GetGainEst.html)
```{bash, eval=F, include=F}
rsync --update --archive --verbose /workdir/marnin/NRCRI_2020GS/ mw489@cbsulm13.biohpc.cornell.edu:/workdir/mw489/NRCRI_2020GS;
rsync --update --archive --verbose /workdir/marnin/NRCRI_2020GS/ mw489@cbsulm15.biohpc.cornell.edu:/workdir/mw489/NRCRI_2020GS;
rsync --update --archive --verbose /workdir/marnin/NRCRI_2020GS/ mw489@cbsulm18.biohpc.cornell.edu:/workdir/mw489/NRCRI_2020GS;
rsync --update --archive --verbose /workdir/mw489/NRCRI_2020GS/ mw489@cbsurobbins.biohpc.cornell.edu:/workdir/marnin/NRCRI_2020GS;
```