Crosses to predict

Preducted untested cross variances

Model A and AD

Set-up
Run Var(TGV) predictions (Model AD)
Run Var(BV) predictions (Model A)

Model DirDom

Set-up
Run Var(TGV) predictions
Run Var(BV) predictions

Re-predict self-crosses

Model A and AD

Predict untested cross means

Set-up
Run mean predictions (Model A and AD)
Run mean predictions (Model DirDom)

Process prediction results

Tidy predicted variances
Tidy predicted means
Compute predictions on SI

Variances
Means

Last updated: 2021-03-24

Checks: 7 0

Knit directory: PredictOutbredCrossVar/

This reproducible R Markdown analysis was created with workflowr (version 1.6.2). The Checks tab describes the reproducibility checks that were applied when the results were created. The Past versions tab lists the development history.

R Markdown file: up-to-date

Great! Since the R Markdown file has been committed to the Git repository, you know the exact version of the code that produced these results.

Environment: empty

Great job! The global environment was empty. Objects defined in the global environment can affect the analysis in your R Markdown file in unknown ways. For reproduciblity it’s best to always run the code in an empty environment.

Seed: set.seed(20191123)

The command set.seed(20191123) was run prior to running the code in the R Markdown file. Setting a seed ensures that any results that rely on randomness, e.g. subsampling or permutations, are reproducible.

Session information: recorded

Great job! Recording the operating system, R version, and package versions is critical for reproducibility.

Cache: none

Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.

File paths: relative

Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.

Repository version: f73b05f

Great! You are using Git for version control. Tracking code development and connecting the code version to the results is critical for reproducibility.

The results in this page were generated with repository version f73b05f. See the Past versions tab to see a history of the changes made to the R Markdown and HTML files.

Note that you need to be careful to ensure that all relevant files for the analysis have been committed to Git prior to generating the results (you can use wflow_publish or wflow_git_commit). workflowr only checks the R Markdown file, but you know if there are other scripts or data files that it depends on. Below is the status of the Git repository when the results were generated:


Ignored files:
    Ignored:    .DS_Store
    Ignored:    .Rhistory
    Ignored:    .Rproj.user/
    Ignored:    output/.DS_Store

Untracked files:
    Untracked:  Icon
    Untracked:  PredictOutbredCrossVarMS_ResponseToReviews_R1.gdoc
    Untracked:  figure/
    Untracked:  manuscript/
    Untracked:  output/crossPredictions/
    Untracked:  output/gblups_DirectionalDom_parentwise_crossVal_folds.rds
    Untracked:  output/gblups_geneticgroups.rds
    Untracked:  output/gblups_parentwise_crossVal_folds.rds
    Untracked:  output/mtMarkerEffects/

Unstaged changes:
    Modified:   analysis/NGCleadersCall.Rmd
    Modified:   code/fitDirectionalDomMtBRR.R
    Modified:   code/fitmtBRR.R
    Modified:   code/getDirectionalDomGenomicBLUPs.R
    Modified:   code/getDirectionalDomMtCrossMeanPreds.R
    Modified:   code/getDirectionalDomMtCrossVarBVpreds.R
    Modified:   code/getDirectionalDomMtCrossVarTGVpreds.R
    Modified:   code/getDirectionalDomVarComps.R
    Modified:   code/getGenomicBLUPs.R
    Modified:   code/getMtCrossMeanPreds.R
    Modified:   code/getMtCrossVarPreds.R
    Modified:   code/getUntestedMtCrossVarPreds.R
    Modified:   code/getVarComps.R
    Modified:   data/blups_forawcdata.rds
    Modified:   data/genmap_awc_May2020.rds
    Modified:   data/parentwise_crossVal_folds.rds
    Modified:   data/ped_awc.rds
    Modified:   data/selection_index_weights_4traits.rds
    Modified:   output/CrossesToPredict_top100stdSI_and_209originalParents.rds
    Modified:   output/accuraciesMeans.rds
    Modified:   output/accuraciesUC.rds
    Modified:   output/accuraciesVars.rds
    Modified:   output/crossRealizations/realizedCrossMeans.rds
    Modified:   output/crossRealizations/realizedCrossMeans_BLUPs.rds
    Modified:   output/crossRealizations/realizedCrossMetrics.rds
    Modified:   output/crossRealizations/realizedCrossVars.rds
    Modified:   output/crossRealizations/realizedCrossVars_BLUPs.rds
    Modified:   output/crossRealizations/realized_cross_means_and_covs_traits.rds
    Modified:   output/crossRealizations/realized_cross_means_and_vars_selindices.rds
    Modified:   output/ddEffects.rds
    Modified:   output/gebvs_ModelA_GroupAll_stdSI.rds
    Modified:   output/obsVSpredMeans.rds
    Modified:   output/obsVSpredUC.rds
    Modified:   output/obsVSpredVars.rds
    Modified:   output/pmv_DirectionalDom_varcomps_geneticgroups.rds
    Modified:   output/pmv_varcomps_geneticgroups.rds
    Modified:   output/pmv_varcomps_geneticgroups_tidy_includingSIvars.rds
    Modified:   output/propHomozygous.rds
    Modified:   output/top100stdSI.rds

Note that any generated files, e.g. HTML, png, CSS, etc., are not included in this status report because it is ok for generated content to have uncommitted changes.

These are the previous versions of the repository in which changes were made to the R Markdown (analysis/predictUntestedCrosses.Rmd) and HTML (docs/predictUntestedCrosses.html) files. If you’ve configured a remote Git repository (see ?wflow_git_remote), click on the hyperlinks in the table below to view the files as they were in that past version.

File	Version	Author	Date	Message
Rmd	f73b05f	wolfemd	2021-03-24	Update to match revised manuscript. Several comparisons moved to Appendix to streamline primary results, figures, etc.
html	4de1330	wolfemd	2021-02-01	Build site.
Rmd	883b1d4	wolfemd	2021-02-01	Update the syntax highlighting and code-block formatting throughout for
Rmd	6a10c30	wolfemd	2021-01-04	Submission and GitHub ready version.
html	6a10c30	wolfemd	2021-01-04	Submission and GitHub ready version.

# devtools::install_github("wolfemd/predCrossVar", ref = 'master', force=T)

Crosses to predict

The 462 crosses actually made
The 100 clones with top rank on the StdSI
The 209 clones already used as parents

rm(list=ls()); gc()
library(tidyverse); library(magrittr); library(predCrossVar); library(BGLR);

# BLUPs -----------
blups<-readRDS(here::here("data","blups_forawcdata.rds")) %>% 
  select(Trait,blups) %>% 
  unnest(blups) %>% 
  select(Trait,germplasmName,drgBLUP) %>% 
  spread(Trait,drgBLUP) %>%  
  select(germplasmName,all_of(c("DM","logFYLD","MCMDS","TCHART"))) # precaution to ensure consistent column order

# Pedigree -----------
ped<-readRDS(here::here("data","ped_awc.rds")) %>%
  distinct(sireID,damID)
parents<-union(ped$sireID,ped$damID)

# Selection weights -----------
indices<-readRDS(file=here::here("data","selection_index_weights_4traits.rds"))

# GEBVs --------------
gebvs<-readRDS(here::here("output","gblups_geneticgroups.rds")) %>% 
  filter(Group=="All",Model=="A") %>% 
  unnest(GBLUPs) %>% 
  select(germplasmName,all_of(c("DM","logFYLD","MCMDS","TCHART")))
gebvMat<-gebvs %>% 
  column_to_rownames(var = "germplasmName") %>% 
  as.matrix
## GEBVs on the "standard" selection index
gebvs %<>% 
  mutate(stdSI=as.numeric(gebvMat%*%indices$stdSI))
saveRDS(gebvs,here::here("output","gebvs_ModelA_GroupAll_stdSI.rds"))
## The top 100 on the index
top100stdSI<-gebvs %>% 
  arrange(desc(stdSI)) %>% 
  slice(1:100) %$% germplasmName
saveRDS(top100stdSI,here::here("output","top100stdSI.rds"))

# table(top100stdSI %in% parents) # only 3
# length(grep("TMS13|TMS14|TMS15",top100stdSI, invert = T)) # 2
# length(grep("TMS13",top100stdSI)) # 52
# length(grep("TMS14",top100stdSI)) # 31
# length(grep("TMS15",top100stdSI)) # 15

## Highest BV -------------
# gebvs %>% 
#   slice_max(order_by = stdSI, n=1) # TMS13F1095P0013
## Lowest BV
# gebvs %>% 
#   filter(germplasmName %in% union(parents,top100stdSI)) %>% 
#   slice_min(order_by = stdSI, n=1) # IITA-TMS-IBA011371

# Crosses To Predict -------------
CrossesToPredict<-crosses2predict(union(parents,top100stdSI)) %>%  # makes df of pairwise non-recprical, selfs-included crosses
  bind_rows(ped %>% # add the crosses already made (for convenience)
              distinct(sireID,damID)) %>% 
  distinct # avoid duplication
# nrow(CrossesToPredict) # [1] 47083
saveRDS(CrossesToPredict,here::here("output","CrossesToPredict_top100stdSI_and_209originalParents.rds"))

Preducted untested cross variances

Model A and AD

Set-up

# activate multithread OpenBLAS
export OMP_NUM_THREADS=88;

library(tidyverse); library(magrittr); library(predCrossVar); library(BGLR);

# Crosses To Predict -------------
CrossesToPredict<-readRDS(here::here("output","CrossesToPredict_top100stdSI_and_209originalParents.rds"))

# Pedigree -----------
ped<-readRDS(here::here("data","ped_awc.rds")) %>%
  distinct(sireID,damID)
parents<-union(ped$sireID,ped$damID)
top100stdSI<-readRDS(here::here("output","top100stdSI.rds"))

# Recomb frequency matrix ------------
recombFreqMat<-readRDS(here::here("data","recombFreqMat_1minus2c_awcmap_May2020.rds"))

# Haplotype Matrix ------------
haploMat<-readRDS(file=here::here("data","haps_awc.rds"))
parenthaps<-sort(c(paste0(union(parents,top100stdSI),"_HapA"),
                   paste0(union(parents,top100stdSI),"_HapB")))
haploMat<-haploMat[parenthaps,colnames(recombFreqMat)]; dim(haploMat)

# Parallelization specs ---------
require(furrr); 
options(future.globals.maxSize=50000*1024^2)
ncores<-10; 

# MCMC params ------
nIter<-30000; burnIn<-5000; thin<-5

# Path for output ----------
outpath<-"output/crossPredictions"

# getUntestedMtCrossVarPreds function -------------
## Function to run for each rep-fold-Model (==unique set of marker effects), predict the relevant cross variances.
source(here::here("code","getUntestedMtCrossVarPreds.R"))

# Divide CrossesToPredict into chunks for each server ------------
nchunks<-5
CrossesToPredict %<>% 
  mutate(Chunk=rep(1:nchunks, each=ceiling(nrow(.)/nchunks), length.out=nrow(.))) %>% 
  nest(data=c(sireID,damID))

Run Var(TGV) predictions (Model AD)

# cbsulm13 - done
chunk<-1;
# cbsulm15 - done
chunk<-2;
# cbsulm20 - done
chunk<-3;
# cbsulm22 - done
chunk<-4;
# cbsulm23 - done
chunk<-5;

# Start run on each server / chunk:  Aug 03 at 6:50AM
getUntestedMtCrossVarPreds(inprefix = "mt_All_AD",
                           outpath = "output/crossPredictions",
                           outprefix = paste0("predUntestedCrossTGVs_chunk",chunk,"_AD"),
                           predType="VPM", Model = "AD", nIter=30000, burnIn=5000,thin=5,
                           CrossesToPredict=CrossesToPredict$data[[chunk]],
                           recombFreqMat=recombFreqMat,haploMat=haploMat,ncores=ncores)

Run Var(BV) predictions (Model A)

# cbsulm13 - aug 4, 10pm - done
chunk<-1;
# cbsulm15 - aug 4, 10pm - done
chunk<-2;
# cbsulm20 - aug 4, 6:55am - done
chunk<-3;
# cbsulm22 - aug 4, 10pm - done 
chunk<-4;
# cbsulm23 - aug 4, 6:55am - done
chunk<-5;

# Start run on each server / chunk: 
getUntestedMtCrossVarPreds(inprefix = "mt_All_A",
                           outpath = "output/crossPredictions",
                           outprefix = paste0("predUntestedCrossBVs_chunk",chunk,"_A"),
                           predType="VPM", Model = "A", nIter=30000, burnIn=5000,thin=5,
                           CrossesToPredict=CrossesToPredict$data[[chunk]],
                           recombFreqMat=recombFreqMat,haploMat=haploMat,ncores=ncores)

Model DirDom

Set-up

# activate multithread OpenBLAS
export OMP_NUM_THREADS=112;

library(tidyverse); library(magrittr); library(predCrossVar); library(BGLR);

# Crosses To Predict -------------
CrossesToPredict<-readRDS(here::here("output","CrossesToPredict_top100stdSI_and_209originalParents.rds"))

# Pedigree -----------
ped<-readRDS(here::here("data","ped_awc.rds")) %>%
  distinct(sireID,damID)
parents<-union(ped$sireID,ped$damID)
top100stdSI<-readRDS(here::here("output","top100stdSI.rds"))

# Recomb frequency matrix ------------
recombFreqMat<-readRDS(here::here("data","recombFreqMat_1minus2c_awcmap_May2020.rds"))

# Haplotype Matrix ------------
haploMat<-readRDS(file=here::here("data","haps_awc.rds"))
parenthaps<-sort(c(paste0(union(parents,top100stdSI),"_HapA"),
                   paste0(union(parents,top100stdSI),"_HapB")))
haploMat<-haploMat[parenthaps,colnames(recombFreqMat)]; dim(haploMat)

# SNP data ------------
snps<-readRDS(here::here("data","dosages_awc.rds")) %>% 
  remove_invariant(.); 

# Parallelization specs ---------
require(furrr); 
options(future.globals.maxSize=50000*1024^2)
ncores<-10; 

# MCMC params ------
nIter<-30000; burnIn<-5000; thin<-5

# Path for output ----------
outpath<-"output/crossPredictions"

# getUntestedMtCrossVarPreds function -------------
## Function to run for each rep-fold-Model (==unique set of marker effects), predict the relevant cross variances.
source(here::here("code","getUntestedMtCrossVarPreds.R"))

# Divide CrossesToPredict into chunks for each server ------------
nchunks<-4
CrossesToPredict %<>% 
  mutate(Chunk=rep(1:nchunks, each=ceiling(nrow(.)/nchunks), length.out=nrow(.))) %>% 
  nest(data=c(sireID,damID))

Run Var(TGV) predictions

# cbsulm13 - Done!
chunk<-1;
# cbsulm17 - Done!
chunk<-2;
# cbsulm12 - Done!
chunk<-3;
# cbsulm16 - Done!
chunk<-4;

# Start run on each server / chunk: 
getDirDomUntestedMtCrossVarTGVpreds(inprefix = "mt_All_DirectionalDom",
                                    outpath = "output/crossPredictions",
                                    outprefix = paste0("predUntestedCrossTGVs_chunk",chunk,"_DirDom"),
                                    predType="VPM", nIter=30000, burnIn=5000,thin=5,
                                    CrossesToPredict=CrossesToPredict$data[[chunk]],
                                    recombFreqMat=recombFreqMat,haploMat=haploMat,ncores=ncores)

Run Var(BV) predictions

# cbsulm26 - Done! (~32hrs)
chunk<-1;
# cbsulm15 - Done!
chunk<-2;
# cbsulm26 - Done!
chunk<-3;
# cbsulm17 - Done!
chunk<-4;

# Start run on each server / chunk: 
getDirDomUntestedMtCrossVarBVpreds(inprefix = "mt_All_DirectionalDom",
                                   outpath = "output/crossPredictions",
                                   outprefix = paste0("predUntestedCrossBVs_chunk",chunk,"_DirDom"),
                                   predType="VPM", nIter=30000, burnIn=5000,thin=5,
                                   CrossesToPredict=CrossesToPredict$data[[chunk]],
                                   recombFreqMat=recombFreqMat,haploMat=haploMat,doseMat=snps,ncores=ncores)

Re-predict self-crosses

Discovered a bug effecting self-crosses. The original version of predCrossVar run (circa July 2020) incorrectly calculated gametic LD matrices with duplicated haplotypes for cases where sireID==damID. Fixed the bug in the package, archived the original version via Git/GitHub. Re-install predCrossVar on server and re-do predictions of selfs.

Model A and AD

Set-up

# activate multithread OpenBLAS
export OMP_NUM_THREADS=88;

library(tidyverse); library(magrittr); library(predCrossVar); library(BGLR);

# Crosses To Predict -------------
CrossesToPredict<-readRDS(here::here("output","CrossesToPredict_top100stdSI_and_209originalParents.rds"))
# just the selfs
CrossesToPredict %<>% filter(sireID==damID)
# Pedigree -----------
ped<-readRDS(here::here("data","ped_awc.rds")) %>%
  distinct(sireID,damID)
parents<-union(ped$sireID,ped$damID)
top100stdSI<-readRDS(here::here("output","top100stdSI.rds"))

# Recomb frequency matrix ------------
recombFreqMat<-readRDS(here::here("data","recombFreqMat_1minus2c_awcmap_May2020.rds"))

# Haplotype Matrix ------------
haploMat<-readRDS(file=here::here("data","haps_awc.rds"))
parenthaps<-sort(c(paste0(union(parents,top100stdSI),"_HapA"),
                   paste0(union(parents,top100stdSI),"_HapB")))
haploMat<-haploMat[parenthaps,colnames(recombFreqMat)]; dim(haploMat)

# Parallelization specs ---------
require(furrr); 
options(future.globals.maxSize=50000*1024^2)
ncores<-10; 

# MCMC params ------
nIter<-30000; burnIn<-5000; thin<-5

# Path for output ----------
outpath<-"output/crossPredictions"

# getUntestedMtCrossVarPreds function -------------
## Function to run for each rep-fold-Model (==unique set of marker effects), predict the relevant cross variances.
source(here::here("code","getUntestedMtCrossVarPreds.R"))

Run Var(TGV) predictions (Model AD)

# Start run on each server / chunk:  
getUntestedMtCrossVarPreds(inprefix = "mt_All_AD",
                           outpath = "output/crossPredictions",
                           outprefix = paste0("predUntestedCrossTGVs_ReDoSelfs_AD"),
                           predType="VPM", Model = "AD", 
                           nIter=30000, burnIn=5000,thin=5,
                           CrossesToPredict=CrossesToPredict,
                           recombFreqMat=recombFreqMat,
                           haploMat=haploMat,ncores=ncores)

Run Var(BV) predictions (Model A)

# Start run on each server / chunk: 
getUntestedMtCrossVarPreds(inprefix = "mt_All_A",
                           outpath = "output/crossPredictions",
                           outprefix = paste0("predUntestedCrossBVs_ReDoSelfs_A"),
                           predType="VPM", Model = "A", 
                           nIter=30000, burnIn=5000,thin=5,
                           CrossesToPredict=CrossesToPredict,
                           recombFreqMat=recombFreqMat,
                           haploMat=haploMat,ncores=ncores)

Predict untested cross means

Set-up

# activate multithread OpenBLAS
export OMP_NUM_THREADS=112;

library(tidyverse); library(magrittr); library(predCrossVar); library(BGLR);
# Crosses To Predict -------------
CrossesToPredict<-readRDS(here::here("output","CrossesToPredict_top100stdSI_and_209originalParents.rds"))

# Pedigree -----------
ped<-readRDS(here::here("data","ped_awc.rds")) %>%
  distinct(sireID,damID)
parents<-union(ped$sireID,ped$damID)
top100stdSI<-readRDS(here::here("output","top100stdSI.rds"))

# SNP data ------------
snps<-readRDS(here::here("data","dosages_awc.rds")) %>% 
  remove_invariant(.); 

# Parallelization specs ---------
require(furrr); 
options(future.globals.maxSize=50000*1024^2)
ncores<-10; 

# MCMC params ------
nIter<-30000; burnIn<-5000; thin<-5

# Path for output ----------
outpath<-"output/crossPredictions"

# getMtCrossMeanPreds function -------------
source(here::here("code","getMtCrossMeanPreds.R"))
# getDirectionalDomMtCrossMeanPreds function -------------
source(here::here("code","getDirectionalDomMtCrossMeanPreds.R"))

# sampleIDs -------
blups<-readRDS(here::here("data","blups_forawcdata.rds")) %>% 
  select(Trait,blups) %>% # BLUPs long-->wide for multivar analysis
  unnest(blups) %>% 
  select(Trait,germplasmName,drgBLUP) %>% 
  spread(Trait,drgBLUP) %>% 
  select(germplasmName,all_of(c("DM","logFYLD","MCMDS","TCHART"))) # precaution to ensure consistent column order
sampleIDs<-blups$germplasmName[blups$germplasmName %in% rownames(snps)]
rm(blups)

Run mean predictions (Model A and AD)

# Done Aug 8
predictedUntestedCrossMeansBV<-getMtCrossMeanPreds(outprefix="mt_All_A",
                                                   Model="A",
                                                   CrossesToPredict=CrossesToPredict,
                                                   doseMat=snps,
                                                   sampleIDs = sampleIDs)
predictedUntestedCrossMeansTGV<-getMtCrossMeanPreds(outprefix="mt_All_AD",
                                                    Model="AD",
                                                    CrossesToPredict=CrossesToPredict,
                                                    doseMat=snps,
                                                    sampleIDs = sampleIDs)

saveRDS(predictedUntestedCrossMeansBV,file=here::here("output/crossPredictions","predictedUntestedCrossMeansBV.rds"))
saveRDS(predictedUntestedCrossMeansTGV,file=here::here("output/crossPredictions","predictedUntestedCrossMeansTGV.rds"))

Run mean predictions (Model DirDom)

predictedUntestedCrossMeansDirDom<-getDirectionalDomMtCrossMeanPreds(outprefix="mt_All_DirectionalDom",
                                                                     CrossesToPredict=CrossesToPredict,
                                                                     doseMat=snps,
                                                                     sampleIDs = sampleIDs)

saveRDS(predictedUntestedCrossMeansDirDom,
        file=here::here("output/crossPredictions","predictedUntestedCrossMeansDirDom.rds"))

Process prediction results

Tidy predicted variances

Next step, used a server, unpacking output for 10 varcomps x ~47K crosses.

Keep the dom. variance in the “tidied” output long enough to calculate dom. variance on the sel. indices. This was previously not done.

Calc. VarTGV=VarA + VarD only for the sup. tables (Tables S17 and S18), which are used for plots and summaries in the manuscript.

library(tidyverse); library(magrittr); library(predCrossVar) 
# Model A
predUntestedCrossVarBVs<-list.files(here::here("output/crossPredictions")) %>% 
  grep("predUntestedCrossBVs",.,value = T) %>% 
  tibble(File=.) %>% 
  mutate(Model=ifelse(grepl("_A_",File),"A","DirDomBV"),
         crossPredictions=map(File,~readRDS(here::here("output/crossPredictions",.)))) %>% 
  unnest_wider(crossPredictions) %>% 
  unnest(varcovars) %>% 
  select(-totcomputetime) %>% 
  unnest_wider(varcomps) # 10 variances/covariances x 5 chunks x 1 models = 50 blocks of ~10K crosses chunk per varcomp

require(furrr); options(mc.cores=50); plan(multiprocess)
predUntestedCrossVarBVs %<>% # in parallel across the 50 chunks
  mutate(predictedfamvars=future_map(predictedfamvars,~unnest(.,predVars) %>% select(-PMV))) %>% 
  select(-totcomputetime) %>% 
  unnest(predictedfamvars) %>% 
  rename(predVar=VPM) %>% 
  mutate(predOf="VarBV")
predUntestedCrossVarBVs %<>% select(-VarComp,-totcomputetime)

library(tidyverse); library(magrittr); library(predCrossVar) 

# Model AD
predUntestedCrossVarTGVs<-list.files(here::here("output/crossPredictions")) %>% 
  grep("predUntestedCrossTGVs",.,value = T) %>% 
  tibble(File=.) %>% 
  mutate(Model=ifelse(grepl("_AD_",File),"AD","DirDomAD"),
         crossPredictions=map(File,~readRDS(here::here("output/crossPredictions",.)))) %>% 
  unnest_wider(crossPredictions) %>% 
  unnest(varcovars) %>% 
  select(-totcomputetime) %>% 
  unnest_wider(varcomps) # 10 variances/covariances x 5 chunks x 2 models = 100 blocks of ~10K crosses chunk per varcomp
predUntestedCrossVarTGVs$predictedfamvars[[1]] %>% 
  mutate(Nsegsnps=map_dbl(predVars,~.$Nsegsnps[1]))
require(furrr); options(mc.cores=50); plan(multiprocess)
predUntestedCrossVarTGVs %<>% # in parallel across the 100 chunks
  mutate(predictedfamvars=future_map(predictedfamvars,function(predictedfamvars){
    return(predictedfamvars %<>% 
             # format each families output in serial 
             mutate(Nsegsnps=map_dbl(predVars,~.$Nsegsnps[1]),
                    predVars=map(predVars,function(predVars){ 
                      return(predVars %>%
                               select(VarComp,VPM)) })) %>% 
             unnest(predVars))})) %>% 
  select(-totcomputetime) %>% 
  unnest(predictedfamvars) %>% 
  rename(predVar=VPM) %>% 
  rename(predOf=VarComp)

Remove the original predictions for selfs and replace with the correct predictions.

# verify nrow before removing "re-do's"
predUntestedCrossVarBVs %>% dim() # [1] 947780      9
# df of re-predicted selfs
## should only be from the ClassicAD model
redoBVs<-predUntestedCrossVarBVs %>% 
  filter(grepl("_ReDoSelfs_",File))
dim(redoBVs) # [1] 6120    9
redoBVs %>% count(Model,predOf)
#   Model predOf     n
#   <chr> <chr>  <int>
# 1 A     VarBV   3060

# df of original predictions (selfs and outcrosses)
# predUntestedCrossVarBVs %>% 
#   filter(!grepl("_ReDoSelfs_",File)) %>% # dim() # [1] 941660      9 # verify redo's removed
#   filter(sireID==damID) %>% # dim() # [1] 6120    9 original self's predictions remain at this point
predUntestedCrossVarBVs %<>% 
  # remove all selfs from the ClassicAD model
  filter(sireID!=damID | (sireID==damID & Model=="DirDomBV")) %>% # dim() # [1] 935540      9
  # add back the corrected predictions for selfs
  bind_rows(.,redoBVs)

redoTGVs<-predUntestedCrossVarTGVs %>% 
  filter(grepl("_ReDoSelfs_",File))
dim(redoTGVs) # [1] 12240   11
predUntestedCrossVarTGVs %<>% 
  # remove all selfs
  filter(sireID!=damID | (sireID==damID & Model=="DirDomAD")) %>% # dim() # [1] 1871080     11
  # add back the corrected predictions for selfs
  bind_rows(.,redoTGVs)
# dim(predUntestedCrossVarTGVs) # [1] 1883320       9

predUntestedCrossVars<-bind_rows(predUntestedCrossVarBVs,
                                 predUntestedCrossVarTGVs)
  # left_join(predUntestedCrossVarBVs %>% 
  #                                  rename(predVarBV=predVarA) %>% 
  #                                  select(-totcomputetime,-File),
  #                                predUntestedCrossVarTGVs %>% 
  #                                  rename(predVarTGV=predVarTot) %>% 
  #                                  select(-predVarA,-predVarD,-totcomputetime,-File)) %>% 
  # pivot_longer(cols=c(predVarBV,predVarTGV),
  #              names_to = "predOf",values_to = "predVar")
# predUntestedCrossVars<-bind_rows(predUntestedCrossVarBVs,
#                                  predUntestedCrossVarTGVs)
#predUntestedCrossVars %<>% select(Model,Trait1,Trait2,sireID,damID,predOf,predVar,Nsegsnps,totcomputetime)

saveRDS(predUntestedCrossVars,here::here("output/crossPredictions","predictedUntestedCrossVars_tidy_traits.rds"))

Tidy predicted means

library(tidyverse); library(magrittr); library(predCrossVar)
predictedUntestedCrossMeansBV<-readRDS(here::here("output/crossPredictions","predictedUntestedCrossMeansBV.rds")) %>%
  pluck("predictedCrossMeans")
predictedUntestedCrossMeansTGV<-readRDS(here::here("output/crossPredictions","predictedUntestedCrossMeansTGV.rds")) %>%
  pluck("predictedCrossMeans")
predictedUntestedCrossMeansDirDom<-readRDS(here::here("output/crossPredictions","predictedUntestedCrossMeansDirDom.rds")) %>%
  pluck("predictedCrossMeans")
predictedUntestedCrossMeans<-predictedUntestedCrossMeansBV %>% 
  left_join(predictedUntestedCrossMeansTGV) %>% 
  mutate(Model="ClassicAD") %>% 
  bind_rows(predictedUntestedCrossMeansDirDom %>% 
              mutate(Model="DirDom"))
saveRDS(predictedUntestedCrossMeans,here::here("output/crossPredictions","predictedUntestedCrossMeans_tidy_traits.rds"))

Compute predictions on SI

Our focus in evaluating predictions of untested crosses will be on predictions on the two SI.

predVarBV - Model A (ClassicAD)
predVarTGV - Model AD (ClassicAD)
predVarBV - DirDomA (DirDom)
predVarTGV - DirDomAD (DirDom)

times

stdSI
biofortSI

for predicted means and variances.

Variances

library(tidyverse); library(magrittr);
# Selection weights -----------
indices<-readRDS(file=here::here("data","selection_index_weights_4traits.rds"))

predictedUntestedCrossVars<-readRDS(here::here("output/crossPredictions","predictedUntestedCrossVars_tidy_traits.rds"))  
#  select(-predVarD,-totcomputetime)

## Predicted Index Variances
predictedUntestedCrossVars_SI<-predictedUntestedCrossVars %>% 
  # filter(Model=="A") %>% 
  # mutate(Model="ClassicAD") %>% 
  # rename(predVarBV=predVarA) %>% 
  # select(-predVarTot) %>% 
  # left_join(predictedUntestedCrossVars %>% 
  #             filter(Model=="ClassicAD") %>% 
  #             rename(predVarTGV=predVarTot) %>% 
  #             select(-predVarA)) %>% 
  # bind_rows(predictedUntestedCrossVars %>% 
  #             filter(Model=="DirDomAD") %>% 
  #             rename(predVarBV=predVarA,
  #                    predVarTGV=predVarTot)) %>% 
  # pivot_longer(cols=c(predVarBV,predVarTGV),names_to = "predOf",values_to = "predVar") %>% 
  nest(varcovars=c(Trait1,Trait2,predVar)) 
require(furrr); options(mc.cores=50); plan(multiprocess)
predictedUntestedCrossVars_SI %<>% 
  mutate(varcovars=future_map(varcovars,
                       function(varcovars){
                         # pairwise to square symmetric matrix
                         gmat<-varcovars %>% 
                           spread(Trait2,predVar) %>% 
                           column_to_rownames(var = "Trait1") %>% 
                           as.matrix %>% 
                           .[indices$Trait,indices$Trait]
                         gmat[lower.tri(gmat)]<-t(gmat)[lower.tri(gmat)]
                         return(gmat) }))

predictedUntestedCrossVars_SI %<>% 
  mutate(stdSI=future_map_dbl(varcovars,~t(indices$stdSI)%*%.%*%indices$stdSI),
         biofortSI=future_map_dbl(varcovars,~t(indices$biofortSI)%*%.%*%indices$biofortSI)) %>% 
  select(-varcovars)
predictedUntestedCrossVars_SI %<>% 
  pivot_longer(cols = c(stdSI,biofortSI),
               names_to = "Trait1", 
               values_to = "predVar") %>% 
  mutate(Trait2=Trait1)
#   pivot_wider(names_from = "predOf", values_from = "predVar")
# predictedUntestedCrossVars_SI %<>% 
#   pivot_longer(cols = c(predVarBV,predVarTGV), names_to = "predOf", values_to = "predVar")

predictedUntestedCrossVars_SI %<>% 
  select(Trait1,Trait2,sireID,damID,Nsegsnps,Model,predOf,predVar)
saveRDS(predictedUntestedCrossVars_SI,here::here("output/crossPredictions","predictedUntestedCrossVars_SelIndices.rds"))

Means

# Selection weights -----------
indices<-readRDS(file=here::here("data","selection_index_weights_4traits.rds"))
predictedUntestedCrossMeans<-readRDS(here::here("output/crossPredictions","predictedUntestedCrossMeans_tidy_traits.rds"))
predictedUntestedCrossMeans %<>% 
  pivot_longer(cols = c(sireGEBV,damGEBV,predMeanBV,predMeanGV), names_to = "predOf", values_to = "predMean") 

## Predicted Index Means
predictedUntestedCrossMeans %<>% 
  spread(Trait,predMean) %>% 
  nest(predMeans=all_of(indices$Trait)) %>% 
  mutate(stdSI=map_dbl(predMeans,~as.matrix(.)%*%indices$stdSI),
         biofortSI=map_dbl(predMeans,~as.matrix(.)%*%indices$biofortSI)) %>% 
  select(-predMeans) %>% 
  pivot_longer(cols = c(stdSI,biofortSI), names_to = "Trait", values_to = "predMean")

predictedUntestedCrossMeans %<>% 
  pivot_wider(names_from = "predOf", values_from = "predMean") %>% 
  select(sireID,damID,Model,Trait,sireGEBV,damGEBV,predMeanBV,predMeanGV)

saveRDS(predictedUntestedCrossMeans,here::here("output/crossPredictions","predictedUntestedCrossMeans_SelIndices.rds"))

sessionInfo()

R version 4.0.3 (2020-10-10)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Big Sur 10.16

Matrix products: default
BLAS:   /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRblas.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/4.0/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] workflowr_1.6.2

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.6        whisker_0.4       knitr_1.31        magrittr_2.0.1   
 [5] R6_2.5.0          rlang_0.4.10      fansi_0.4.2       stringr_1.4.0    
 [9] tools_4.0.3       xfun_0.22         utf8_1.2.1        git2r_0.28.0     
[13] jquerylib_0.1.3   htmltools_0.5.1.1 ellipsis_0.3.1    rprojroot_2.0.2  
[17] yaml_2.2.1        digest_0.6.27     tibble_3.1.0      lifecycle_1.0.0  
[21] crayon_1.4.1      later_1.1.0.1     sass_0.3.1        vctrs_0.3.6      
[25] promises_1.2.0.1  fs_1.5.0          glue_1.4.2        evaluate_0.14    
[29] rmarkdown_2.7     stringi_1.5.3     bslib_0.2.4       compiler_4.0.3   
[33] pillar_1.5.1      jsonlite_1.7.2    httpuv_1.5.5      pkgconfig_2.0.3

Predict Untested Cross Means and Variances

2020-July-30

Crosses to predict

Preducted untested cross variances

Model A and AD

Set-up

Run Var(TGV) predictions (Model AD)

Run Var(BV) predictions (Model A)

Model DirDom

Set-up

Run Var(TGV) predictions

Run Var(BV) predictions

Re-predict self-crosses

Model A and AD

Set-up

Run Var(TGV) predictions (Model AD)

Run Var(BV) predictions (Model A)

Predict untested cross means

Set-up

Run mean predictions (Model A and AD)

Run mean predictions (Model DirDom)

Process prediction results

Tidy predicted variances

Tidy predicted means

Compute predictions on SI

Variances

Means