Last updated: 2020-12-04

Checks: 7 0

Knit directory: IITA_2020GS/

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(20200915)

The command set.seed(20200915) 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: 79b6430

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 79b6430. 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:    analysis/.DS_Store
    Ignored:    data/.DS_Store
    Ignored:    output/.DS_Store

Untracked files:
    Untracked:  data/GEBV_IITA_OutliersRemovedTRUE_73119.csv
    Untracked:  data/PedigreeGeneticGainCycleTime_aafolabi_01122020.csv
    Untracked:  data/iita_blupsForCrossVal_outliersRemoved_73019.rds
    Untracked:  output/DosageMatrix_IITA_2020Sep16.rds
    Untracked:  output/IITA_CleanedTrialData_2020Dec03.rds
    Untracked:  output/IITA_ExptDesignsDetected_2020Dec03.rds
    Untracked:  output/Kinship_AA_IITA_2020Sep16.rds
    Untracked:  output/Kinship_AD_IITA_2020Sep16.rds
    Untracked:  output/Kinship_A_IITA_2020Sep16.rds
    Untracked:  output/Kinship_DD_IITA2020Sep16.rds
    Untracked:  output/Kinship_D_IITA_2020Sep16.rds
    Untracked:  output/cvresults_ModelADE_chunk1.rds
    Untracked:  output/cvresults_ModelADE_chunk2.rds
    Untracked:  output/cvresults_ModelADE_chunk3.rds
    Untracked:  output/genomicPredictions_ModelADE_threestage_IITA_2020Sep21.rds
    Untracked:  output/genomicPredictions_ModelADE_twostage_IITA_2020Dec03.rds
    Untracked:  output/genomicPredictions_ModelA_threestage_IITA_2020Sep21.rds
    Untracked:  output/iita_blupsForModelTraining_twostage_asreml_2020Dec03.rds
    Untracked:  output/model_meangetgvs_vs_year.csv
    Untracked:  output/model_rawgetgvs_vs_year.csv
    Untracked:  output/training_data_summary.csv
    Untracked:  workflowr_log.R

Unstaged changes:
    Modified:   output/IITA_ExptDesignsDetected.rds
    Modified:   output/iita_blupsForModelTraining.rds
    Modified:   output/meanGETGVbyYear_IITA_2020Dec03.csv

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/03-GetBLUPs.Rmd) and HTML (docs/03-GetBLUPs.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
html	7bae38d	wolfemd	2020-12-03	Build site.
html	b9bb6f8	wolfemd	2020-12-03	Build site.
html	d72a9ed	wolfemd	2020-09-21	Build site.
html	9194239	wolfemd	2020-09-21	Build site.
Rmd	97778e7	wolfemd	2020-09-21	Big update. Two types of pipeline to get BLUPs, GEBVs and GETGVs:
html	d6d72f8	wolfemd	2020-09-17	Build site.
html	7e156dd	wolfemd	2020-09-17	Build site.
Rmd	7ea8b80	wolfemd	2020-09-17	All steps including genomic predicting (excluding cross-validation),

Previous step

Curate by trait-trial: Model each trait-trial separately, remove outliers, get BLUPs.

This step

Three-stage procedure:

Previous step: Fit mixed-model and extract BLUPs, de-regressed BLUPs and weights on a per-trial basis with one round of outlier removal x1 (as in the previous section)
This step: Fit mixed-model with de-regressed BLUPs and weights from previous step as input. Get BLUPs from multi-trial analysis.
Genomic prediction with drg-BLUPs from multi-trial analysis as input.

Get multi-trial BLUPs from per-trial BLUPs (three-stage)

Set-up training datasets

This next step fits models to each trait, combining curated data (BLUPs) from each trial, which we computed in the previous step.

rm(list = ls())
library(tidyverse)
library(magrittr)
source(here::here("code", "gsFunctions.R"))
dbdata <- readRDS(here::here("output", "IITA_CuratedTrials.rds"))
traits <- c("MCMDS", "DM", "PLTHT", "BRNHT1", "BRLVLS", "HI", "logFYLD", "logTOPYLD", 
    "logRTNO", "TCHART", "LCHROMO", "ACHROMO", "BCHROMO")

Starting with the curated trial data (which correspond to per-trait, per-trial BLUPs) from the previous step.

Nest by trait. Need to restructure the data from per-trial BLUPs by regrouping by trait.

dbdata <- nestForMultiTrialAnalysis(dbdata)

dbdata %>% mutate(N_blups = map_dbl(MultiTrialTraitData, nrow)) %>% rmarkdown::paged_table()

Model multiple trials

Function fitMultiTrialModel() takes de-regressed BLUPs as response and corresponding weights on error variances are applied. Output includes BLUPs for each clone that combine data across trials and are suitable for downstream genomic prediction work.

Apply the fitMultiTrialModel() to each chunk of trials (per trait) using the purrr function map().

dbdata %<>% mutate(modelOutput = map(MultiTrialTraitData, fitMultiTrialModel))

dbdata %>% select(-MultiTrialTraitData) %>% unnest(modelOutput) %>% unnest(VarComps) %>% 
    rmarkdown::paged_table()

Output file

saveRDS(dbdata, file = here::here("output", "iita_blupsForModelTraining.rds"))

Get multi-trial BLUPs from raw data (two-stage)

Between the July 2019 and current genomic evaluation for IITA, I made “upgrades” (?) to the prediction procedure. Instead of the two-stage procedure implemented in 2019, I used a three-stage procedure in 2020 (described at the top). This was previously implemented for the NRCRI prediction done in April 2020.

I think it is essential to compare the “three-stage” approach with the “two-stage”!

Two-stage procedure:

Fit mixed-model to multi-trial dataset and extract BLUPs, de-regressed BLUPs and weights. Include two rounds of outlier removal.
Genomic prediction with drg-BLUPs from multi-trial analysis as input.

Work below represents Stage 1 of the Two-stage procedure.

set-up training datasets

# activate multithread OpenBLAS for fast matrix algebra
export OMP_NUM_THREADS=56

rm(list = ls())
library(tidyverse)
library(magrittr)
# source(here::here('code','gsFunctions.R'))
dbdata <- readRDS(here::here("output", "IITA_ExptDesignsDetected.rds"))
traits <- c("MCMDS", "DM", "PLTHT", "BRNHT1", "BRLVLS", "HI", "logFYLD", "logTOPYLD", 
    "logRTNO", "TCHART", "LCHROMO", "ACHROMO", "BCHROMO")

Nest by trait. Need to restructure the data from per-trial by regrouping by trait.

dbdata %<>% dplyr::select(-MaxNOHAV) %>% unnest(TrialData) %>% dplyr::select(programName, 
    locationName, studyYear, TrialType, studyName, CompleteBlocks, IncompleteBlocks, 
    yearInLoc, trialInLocYr, repInTrial, blockInRep, observationUnitDbId, germplasmName, 
    FullSampleName, GID, all_of(traits), PropNOHAV) %>% mutate(IncompleteBlocks = ifelse(IncompleteBlocks == 
    TRUE, "Yes", "No"), CompleteBlocks = ifelse(CompleteBlocks == TRUE, "Yes", "No")) %>% 
    pivot_longer(cols = all_of(traits), names_to = "Trait", values_to = "Value") %>% 
    filter(!is.na(Value), !is.na(GID)) %>% nest(MultiTrialTraitData = c(-Trait))

To fit the mixed-model I used last year, I am again resorting to asreml. I fit random effects for rep and block only where complete and incomplete blocks, respectively are indicated in the trial design variables. sommer should be able to fit the same model via the at() function, but I am having trouble with it and sommer is much slower even without a dense covariance (i.e. a kinship), compared to lme4::lmer() or asreml().

dbdata %<>% mutate(fixedFormula = ifelse(Trait %in% c("logFYLD", "logRTNO", "logTOPYLD"), 
    "Value ~ yearInLoc", "Value ~ yearInLoc + PropNOHAV"), randFormula = paste0("~idv(GID) + idv(trialInLocYr) + at(CompleteBlocks,'Yes'):repInTrial ", 
    "+ at(IncompleteBlocks,'Yes'):blockInRep"))
dbdata %>% mutate(Nobs = map_dbl(MultiTrialTraitData, nrow)) %>% select(Trait, Nobs, 
    fixedFormula, randFormula) %>% rmarkdown::paged_table()

# randFormula<-paste0('~vs(GID) + vs(trialInLocYr) +
# vs(at(CompleteBlocks,'Yes'),repInTrial) +
# vs(at(IncompleteBlocks,'Yes'),blockInRep)') library(sommer) fit <- mmer(fixed =
# Value ~ 1 + yearInLoc, random = as.formula(randFormula), data=trainingdata,
# getPEV=TRUE)

Function to run asreml

Includes rounds of outlier removal and re-fitting.

fitASfunc<-function(fixedFormula,randFormula,MultiTrialTraitData,...){
  # test arguments for function
  # ----------------------
  # MultiTrialTraitData<-dbdata$MultiTrialTraitData[[7]]
  # #Trait<-dbdata$Trait[[3]]
  # fixedFormula<-dbdata$fixedFormula[[7]]
  # randFormula<-dbdata$randFormula[[7]]
  # test<-fitASfunc(fixedFormula,randFormula,MultiTrialTraitData)
  # ----------------------
  require(asreml); 
  fixedFormula<-as.formula(fixedFormula)
  randFormula<-as.formula(randFormula)
  # fit asreml 
  out<-asreml(fixed = fixedFormula,
              random = randFormula,
              data = MultiTrialTraitData, 
              maxiter = 40, workspace=800e6, na.method.X = "omit")
  #### extract residuals - Round 1
  
  outliers1<-which(abs(scale(out$residuals))>3.3)
  
  if(length(outliers1)>0){
    
    x<-MultiTrialTraitData[-outliers1,]
    # re-fit
    out<-asreml(fixed = fixedFormula,
                random = randFormula,
                data = x, 
                maxiter = 40, workspace=800e6, na.method.X = "omit")
    #### extract residuals - Round 2
    outliers2<-which(abs(scale(out$residuals))>3.3)
    if(length(outliers2)>0){
      #### remove outliers
      x<-x[-outliers2,]
      # final re-fit
      out<-asreml(fixed = fixedFormula,
                  random = randFormula,
                  data = x, maxiter = 40,workspace=800e6, na.method.X = "omit")
    }
  }
  if(length(outliers1)==0){ outliers1<-NULL }
  if(length(outliers2)==0){ outliers2<-NULL }
  
  ll<-summary(out,all=T)$loglik
  varcomp<-summary(out,all=T)$varcomp
  Vg<-varcomp["GID!GID.var","component"]
  Ve<-varcomp["R!variance","component"]
  H2=Vg/(Vg+Ve)
  blups<-summary(out,all=T)$coef.random %>%
    as.data.frame %>%
    rownames_to_column(var = "GID") %>%
    dplyr::select(GID,solution,`std error`) %>%
    filter(grepl("GID",GID)) %>%
    rename(BLUP=solution) %>%
    mutate(GID=gsub("GID_","",GID),
           PEV=`std error`^2, # asreml specific
           REL=1-(PEV/Vg), # Reliability
           drgBLUP=BLUP/REL, # deregressed BLUP
           WT=(1-H2)/((0.1 + (1-REL)/REL)*H2)) # weight for use in Stage 2
  out<-tibble(loglik=ll,Vg,Ve,H2,
              blups=list(blups),
              varcomp=list(varcomp),
              outliers1=list(outliers1),
              outliers2=list(outliers2))
  return(out) }

Run asreml

library(furrr)
options(mc.cores = 13)
plan(multiprocess)
library(asreml)
dbdata %<>% mutate(fitAS = future_pmap(., fitASfunc))
dbdata %<>% select(-fixedFormula, -randFormula, -MultiTrialTraitData) %>% unnest(fitAS)

Output file

saveRDS(dbdata, file = here::here("output", "iita_blupsForModelTraining_twostage_asreml.rds"))

Next step

Check prediction accuracy: Evaluate prediction accuracy with cross-validation.

sessionInfo()

R version 4.0.2 (2020-06-22)
Platform: x86_64-apple-darwin17.0 (64-bit)
Running under: macOS Catalina 10.15.7

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] lme4_1.1-26     Matrix_1.2-18   magrittr_2.0.1  forcats_0.5.0  
 [5] stringr_1.4.0   dplyr_1.0.2     purrr_0.3.4     readr_1.4.0    
 [9] tidyr_1.1.2     tibble_3.0.4    ggplot2_3.3.2   tidyverse_1.3.0
[13] workflowr_1.6.2

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.5        lubridate_1.7.9.2 here_1.0.0        lattice_0.20-41  
 [5] ps_1.4.0          assertthat_0.2.1  rprojroot_2.0.2   digest_0.6.27    
 [9] R6_2.5.0          cellranger_1.1.0  backports_1.2.0   reprex_0.3.0     
[13] evaluate_0.14     httr_1.4.2        pillar_1.4.7      rlang_0.4.9      
[17] readxl_1.3.1      minqa_1.2.4       rstudioapi_0.13   nloptr_1.2.2.2   
[21] whisker_0.4       rmarkdown_2.5     splines_4.0.2     statmod_1.4.35   
[25] munsell_0.5.0     broom_0.7.2       compiler_4.0.2    httpuv_1.5.4     
[29] modelr_0.1.8      xfun_0.19         pkgconfig_2.0.3   htmltools_0.5.0  
[33] tidyselect_1.1.0  fansi_0.4.1       crayon_1.3.4      dbplyr_2.0.0     
[37] withr_2.3.0       later_1.1.0.1     MASS_7.3-53       grid_4.0.2       
[41] nlme_3.1-150      jsonlite_1.7.1    gtable_0.3.0      lifecycle_0.2.0  
[45] DBI_1.1.0         git2r_0.27.1      formatR_1.7       scales_1.1.1     
[49] cli_2.2.0         stringi_1.5.3     fs_1.5.0          promises_1.1.1   
[53] xml2_1.3.2        ellipsis_0.3.1    generics_0.1.0    vctrs_0.3.5      
[57] boot_1.3-25       tools_4.0.2       glue_1.4.2        hms_0.5.3        
[61] yaml_2.2.1        colorspace_2.0-0  rvest_0.3.6       knitr_1.30       
[65] haven_2.3.1

Get BLUPs combining all trial data

wolfemd

2020-Sep-16

Previous step

This step

Get multi-trial BLUPs from per-trial BLUPs (three-stage)

Set-up training datasets

Model multiple trials

Output file

Get multi-trial BLUPs from raw data (two-stage)

set-up training datasets

Function to run asreml

Run asreml

Output file

Next step