Last updated: 2021-08-09
Checks: 7 0
Knit directory: IITA_2021GS/
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.
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.
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.
The command set.seed(20210504) 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.
Great job! Recording the operating system, R version, and package versions is critical for reproducibility.
Nice! There were no cached chunks for this analysis, so you can be confident that you successfully produced the results during this run.
Great job! Using relative paths to the files within your workflowr project makes it easier to run your code on other machines.
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 6f2057f. 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: IITA_2021GS.Rproj
Untracked: code/convertDart2vcf.R
Untracked: code/imputationFunctions.R
Untracked: data/DatabaseDownload_2021Aug08/
Untracked: data/DatabaseDownload_2021May04/
Untracked: data/GBSdataMasterList_31818.csv
Untracked: data/IITA_GBStoPhenoMaster_33018.csv
Untracked: data/NRCRI_GBStoPhenoMaster_40318.csv
Untracked: data/PedigreeGeneticGainCycleTime_aafolabi_01122020.xls
Untracked: data/Report-DCas21-6038/
Untracked: data/chr1_RefPanelAndGSprogeny_ReadyForGP_72719.fam
Untracked: output/IITA_CleanedTrialData_2021Aug08.rds
Untracked: output/IITA_trials_NOT_identifiable.csv
Untracked: output/maxNOHAV_byStudy.csv
Unstaged changes:
Modified: analysis/05-CrossValidation.Rmd
Modified: analysis/06-GenomicPredictions.Rmd
Modified: analysis/07-Results.Rmd
Deleted: analysis/inputsForSimulation.Rmd
Deleted: implementGMSinCassava.Rproj
Deleted: output/README.md
Deleted: output/cvAD_5rep5fold_predAccuracy.rds
Deleted: output/cvDirDom_5rep5fold_predAccuracy.rds
Deleted: output/estimateSelectionError.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/01-cleanTPdata.Rmd) and HTML (docs/01-cleanTPdata.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 | 6f2057f | wolfemd | 2021-08-09 | Publish project. Imputation completed. Run and complete ‘cleanTPdata’ step. |
| html | 934141c | wolfemd | 2021-07-14 | Build site. |
| html | e66bdad | wolfemd | 2021-06-10 | Build site. |
| Rmd | a8452ba | wolfemd | 2021-06-10 | Initial build of the entire page upon completion of all |
Follow outlined GenomicPredictionChecklist and previous pipeline to process cassavabase data for ultimate genomic prediction.
Below we will clean and format training data.
Downloaded all IITA field trials.
Selected all IITA trials currently available. Make a list.
Go to Manage –> Download here. Download phenotypes (plot-basis only) and meta-data as CSV using the corresponding boxes / drop-downs.
Store flatfiles, unaltered in directory data/DatabaseDownload_2021May04/ and data/DatabaseDownload_2021Aug08/.
May 4 download contained all IITA data to date. Trial list is named ALL_IITA_TRIALS_2021May04.
Aug 8 wanted to up date last 3 months of trial uploads. Downloaded 2020 and 2021 data only, using “Wizard” instead of “Download” page. Trial list is named IITA_Trials2020and2021_2021Aug088.
library(tidyverse); library(magrittr);
library(genomicMateSelectR)Read DB data.
dbdata_may4<-readDBdata(phenotypeFile = here::here("data/DatabaseDownload_2021May04"
,"2021-05-04T193557phenotype_download.csv"),
metadataFile = here::here("data/DatabaseDownload_2021May04"
,"2021-05-04T194847metadata_download.csv"))NOTE: Frustratingly, the usual download method for DB data I use (the download page) is non-responsive at the moment. As an alternative, I downloaded via the Wizard instead. The two tools produce slightly different results which breaks my code. Below, manually read and combine the May 4 DB data with Aug 8 download. Replace all trials in the May 4 download with trials data downloaded Aug 8.
indata<-read.csv(here::here("data/DatabaseDownload_2021Aug08"
,"phenotype.csv"),skip = 3,
na.strings = c("#VALUE!",NA,".",""," ","-","\""),
stringsAsFactors = F)
meta<-read.csv(here::here("data/DatabaseDownload_2021Aug08",
"metadata.csv"),skip = 2,
na.strings = c("#VALUE!",NA,".",""," ","-","\""),
stringsAsFactors = F) %>%
rename(programName=breedingProgramName,
programDescription=breedingProgramDescription,
programDbId=breedingProgramDbId)
indata<-left_join(indata,meta)
table(unique(indata$studyName) %in% dbdata_may4$studyName)
dbdata<-dbdata_may4 %>%
filter(!studyName %in% indata$studyName) %>%
mutate(across(contains("CO_334"),~as.numeric(.))) %>%
bind_rows(indata %>%
mutate(replicate=as.character(replicate),
rowNumber=as.character(rowNumber)) %>%
mutate(across(contains("CO_334"),~as.numeric(.))))Make TrialType Variable
dbdata<-makeTrialTypeVar(dbdata)
dbdata %>%
count(TrialType) %>% rmarkdown::paged_table()Looking at the studyName’s of trials getting NA for TrialType, which can’t be classified at present.
Here is the list of trials I am not including.
dbdata %>% filter(is.na(TrialType)) %$% unique(studyName) %>%
write.csv(.,file = here::here("output","IITA_trials_NOT_identifiable.csv"), row.names = F)Wrote to disk a CSV in the output/ sub-directory.
Should any of these trials have been included?
Especially among the following new trials (post 2018):
dbdata %>%
filter(is.na(TrialType),
as.numeric(studyYear)>2018) %$% unique(studyName)dbdata %<>%
filter(!is.na(TrialType))
dbdata %>%
group_by(programName) %>%
summarize(N=n()) %>% rmarkdown::paged_table()
# Aug 2021: 547203
# May 2021: 524390 (now including a ~200K plot seedling nursery) plots
## Dec 2020: 475097 plots (~155K are seedling nurseries which will be excluded from most analyses)Making a table of abbreviations for renaming. Since July 2019 version: added chromometer traits (L, a, b) and added branching levels count (BRLVLS) at IYR’s request.
traitabbrevs<-tribble(~TraitAbbrev,~TraitName,
"CMD1S","cassava.mosaic.disease.severity.1.month.evaluation.CO_334.0000191",
"CMD3S","cassava.mosaic.disease.severity.3.month.evaluation.CO_334.0000192",
"CMD6S","cassava.mosaic.disease.severity.6.month.evaluation.CO_334.0000194",
"CMD9S","cassava.mosaic.disease.severity.9.month.evaluation.CO_334.0000193",
"CGM","Cassava.green.mite.severity.CO_334.0000033",
"CGMS1","cassava.green.mite.severity.first.evaluation.CO_334.0000189",
"CGMS2","cassava.green.mite.severity.second.evaluation.CO_334.0000190",
"DM","dry.matter.content.percentage.CO_334.0000092",
"PLTHT","plant.height.measurement.in.cm.CO_334.0000018",
"BRNHT1","first.apical.branch.height.measurement.in.cm.CO_334.0000106",
"BRLVLS","branching.level.counting.CO_334.0000079",
"SHTWT","fresh.shoot.weight.measurement.in.kg.per.plot.CO_334.0000016",
"RTWT","fresh.storage.root.weight.per.plot.CO_334.0000012",
"RTNO","root.number.counting.CO_334.0000011",
"TCHART","total.carotenoid.by.chart.1.8.CO_334.0000161",
"LCHROMO","L.chromometer.value.CO_334.0002065",
"ACHROMO","a.chromometer.value.CO_334.0002066",
"BCHROMO","b.chromometer.value.CO_334.0002064",
"NOHAV","plant.stands.harvested.counting.CO_334.0000010")
traitabbrevs %>% rmarkdown::paged_table()Run function renameAndSelectCols() to rename columns and remove everything unecessary
dbdata<-renameAndSelectCols(traitabbrevs,indata=dbdata,customColsToKeep = c("TrialType","observationUnitName"))dbdata<-dbdata %>%
mutate(CMD1S=ifelse(CMD1S<1 | CMD1S>5,NA,CMD1S),
CMD3S=ifelse(CMD3S<1 | CMD3S>5,NA,CMD3S),
CMD6S=ifelse(CMD6S<1 | CMD6S>5,NA,CMD6S),
CMD9S=ifelse(CMD9S<1 | CMD9S>5,NA,CMD9S),
CGM=ifelse(CGM<1 | CGM>5,NA,CGM),
CGMS1=ifelse(CGMS1<1 | CGMS1>5,NA,CGMS1),
CGMS2=ifelse(CGMS2<1 | CGMS2>5,NA,CGMS2),
DM=ifelse(DM>100 | DM<=0,NA,DM),
RTWT=ifelse(RTWT==0 | NOHAV==0 | is.na(NOHAV),NA,RTWT),
SHTWT=ifelse(SHTWT==0 | NOHAV==0 | is.na(NOHAV),NA,SHTWT),
RTNO=ifelse(RTNO==0 | NOHAV==0 | is.na(NOHAV),NA,RTNO),
NOHAV=ifelse(NOHAV==0,NA,NOHAV),
NOHAV=ifelse(NOHAV>42,NA,NOHAV),
RTNO=ifelse(!RTNO %in% 1:10000,NA,RTNO))dbdata<-dbdata %>%
mutate(HI=RTWT/(RTWT+SHTWT))I anticipate this will not be necessary as it will be computed before or during data upload.
For calculating fresh root yield:
dbdata<-dbdata %>%
mutate(PlotSpacing=ifelse(programName!="IITA",1,
ifelse(studyYear<2013,1,
ifelse(TrialType %in% c("CET","GeneticGain","ExpCET"),1,0.8))))
maxNOHAV_byStudy<-dbdata %>%
group_by(programName,locationName,studyYear,studyName,studyDesign) %>%
summarize(MaxNOHAV=max(NOHAV, na.rm=T)) %>%
ungroup() %>%
mutate(MaxNOHAV=ifelse(MaxNOHAV=="-Inf",NA,MaxNOHAV))
write.csv(maxNOHAV_byStudy %>% arrange(studyYear),file=here::here("output","maxNOHAV_byStudy.csv"), row.names = F)# I log transform yield traits
# to satisfy homoskedastic residuals assumption
# of linear mixed models
dbdata<-left_join(dbdata,maxNOHAV_byStudy) %>%
mutate(RTWT=ifelse(NOHAV>MaxNOHAV,NA,RTWT),
SHTWT=ifelse(NOHAV>MaxNOHAV,NA,SHTWT),
RTNO=ifelse(NOHAV>MaxNOHAV,NA,RTNO),
HI=ifelse(NOHAV>MaxNOHAV,NA,HI),
FYLD=RTWT/(MaxNOHAV*PlotSpacing)*10,
DYLD=FYLD*(DM/100),
logFYLD=log(FYLD),
logDYLD=log(DYLD),
logTOPYLD=log(SHTWT/(MaxNOHAV*PlotSpacing)*10),
logRTNO=log(RTNO),
PropNOHAV=NOHAV/MaxNOHAV)
# remove non transformed / per-plot (instead of per area) traits
dbdata %<>% select(-RTWT,-SHTWT,-RTNO,-FYLD,-DYLD)# [NEW AS OF APRIL 2021]
## VERSION with vs. without CBSD
## Impervious to particular timepoints between 1, 3, 6 and 9 scores
# Without CBSD (West Africa)
dbdata<-dbdata %>%
mutate(MCMDS=rowMeans(.[,colnames(.) %in% c("CMD1S","CMD3S","CMD6S","CMD9S")], na.rm = T)) %>%
select(-any_of(c("CMD1S","CMD3S","CMD6S","CMD9S")))
# With CBSD (East Africa)
# dbdata<-dbdata %>%
# mutate(MCMDS=rowMeans(.[,colnames(.) %in% c("CMD1S","CMD3S","CMD6S","CMD9S")], na.rm = T),
# MCBSDS=rowMeans(.[,colnames(.) %in% c("CBSD1S","CBSD3S","CBSD6S","CBSD9S")], na.rm = T)) %>%
# select(-any_of(c("CMD1S","CMD3S","CMD6S","CMD9S","CBSD1S","CBSD3S","CBSD6S","CBSD9S")))This step is mostly copy-pasted from previous processing of IITA- and IITA-specific data.
Uses 4 flat files, which are available e.g. here. Specifically, IITA_GBStoPhenoMaster_33018.csv, GBSdataMasterList_31818.csv and IITA_GBStoPhenoMaster_40318.csv and chr1_RefPanelAndGSprogeny_ReadyForGP_72719.fam. I copy them to the data/ sub-directory for the current analysis.
In addition, DArT-only samples are now expected to also have phenotypes. Therefore, checking for matches in new flatfiles, deposited in the data/ (see code below).
library(tidyverse); library(magrittr)
gbs2phenoMaster<-dbdata %>%
select(germplasmName) %>%
distinct %>%
left_join(read.csv(here::here("data","NRCRI_GBStoPhenoMaster_40318.csv"),
stringsAsFactors = F)) %>%
mutate(FullSampleName=ifelse(grepl("C2a",germplasmName,ignore.case = T) &
is.na(FullSampleName),germplasmName,FullSampleName)) %>%
filter(!is.na(FullSampleName)) %>%
select(germplasmName,FullSampleName) %>%
bind_rows(dbdata %>%
select(germplasmName) %>%
distinct %>%
left_join(read.csv(here::here("data","IITA_GBStoPhenoMaster_33018.csv"),
stringsAsFactors = F)) %>%
filter(!is.na(FullSampleName)) %>%
select(germplasmName,FullSampleName)) %>%
bind_rows(dbdata %>%
select(germplasmName) %>%
distinct %>%
left_join(read.csv(here::here("data","GBSdataMasterList_31818.csv"),
stringsAsFactors = F) %>%
select(DNASample,FullSampleName) %>%
rename(germplasmName=DNASample)) %>%
filter(!is.na(FullSampleName)) %>%
select(germplasmName,FullSampleName)) %>%
bind_rows(dbdata %>%
select(germplasmName) %>%
distinct %>%
mutate(germplasmSynonyms=ifelse(grepl("^UG",germplasmName,ignore.case = T),
gsub("UG","Ug",germplasmName),germplasmName)) %>%
left_join(read.csv(here::here("data","GBSdataMasterList_31818.csv"),
stringsAsFactors = F) %>%
select(DNASample,FullSampleName) %>%
rename(germplasmSynonyms=DNASample)) %>%
filter(!is.na(FullSampleName)) %>%
select(germplasmName,FullSampleName)) %>%
bind_rows(dbdata %>%
select(germplasmName) %>%
distinct %>%
mutate(germplasmSynonyms=ifelse(grepl("^TZ",germplasmName,
ignore.case = T),
gsub("TZ","",germplasmName),germplasmName)) %>%
left_join(read.csv(here::here("data","GBSdataMasterList_31818.csv"),
stringsAsFactors = F) %>%
select(DNASample,FullSampleName) %>%
rename(germplasmSynonyms=DNASample)) %>%
filter(!is.na(FullSampleName)) %>%
select(germplasmName,FullSampleName)) %>%
distinct %>%
left_join(read.csv(here::here("data","GBSdataMasterList_31818.csv"),
stringsAsFactors = F) %>%
select(FullSampleName,OrigKeyFile,Institute) %>%
rename(OriginOfSample=Institute)) %>%
mutate(OrigKeyFile=ifelse(grepl("C2a",germplasmName,ignore.case = T),
ifelse(is.na(OrigKeyFile),"LavalGBS",OrigKeyFile),
OrigKeyFile),
OriginOfSample=ifelse(grepl("C2a",germplasmName,ignore.case = T),
ifelse(is.na(OriginOfSample),"NRCRI",OriginOfSample),
OriginOfSample))
## NEW: check for germName-DArT name matches
germNamesWithoutGBSgenos<-dbdata %>%
select(programName,germplasmName) %>%
distinct %>%
left_join(gbs2phenoMaster) %>%
filter(is.na(FullSampleName)) %>%
select(-FullSampleName)
## NEW: check for germName-DArT name matches
germNamesWithoutGBSgenos<-dbdata %>%
select(programName,germplasmName) %>%
distinct %>%
left_join(gbs2phenoMaster) %>%
filter(is.na(FullSampleName)) %>%
select(-FullSampleName)
germNamesWithDArT<-germNamesWithoutGBSgenos %>%
inner_join(read.table(here::here("data","chr1_RefPanelAndGSprogeny_ReadyForGP_72719.fam"),
header = F, stringsAsFactors = F)$V2 %>%
grep("TMS16|TMS17|TMS18|TMS19|TMS20",.,value = T, ignore.case = T) %>%
tibble(dartName=.) %>%
separate(dartName,c("germplasmName","dartID"),"_",extra = 'merge',remove = F)) %>%
group_by(germplasmName) %>%
slice(1) %>%
ungroup() %>%
rename(FullSampleName=dartName) %>%
mutate(OrigKeyFile="DArTseqLD", OriginOfSample="IITA") %>%
select(-dartID)
print(paste0(nrow(germNamesWithDArT)," germNames with DArT-only genos"))
# first, filter to just program-DNAorigin matches
germNamesWithGenos<-dbdata %>%
select(programName,germplasmName) %>%
distinct %>%
left_join(gbs2phenoMaster) %>%
filter(!is.na(FullSampleName))
print(paste0(nrow(germNamesWithGenos)," germNames with GBS genos"))
# program-germNames with locally sourced GBS samples
germNamesWithGenos_HasLocalSourcedGBS<-germNamesWithGenos %>%
filter(programName==OriginOfSample) %>%
select(programName,germplasmName) %>%
semi_join(germNamesWithGenos,.) %>%
group_by(programName,germplasmName) %>% # select one DNA per germplasmName per program
slice(1) %>% ungroup()
print(paste0(nrow(germNamesWithGenos_HasLocalSourcedGBS)," germNames with local GBS genos"))
# the rest (program-germNames) with GBS but coming from a different breeding program
germNamesWithGenos_NoLocalSourcedGBS<-germNamesWithGenos %>%
filter(programName==OriginOfSample) %>%
select(programName,germplasmName) %>%
anti_join(germNamesWithGenos,.) %>%
# select one DNA per germplasmName per program
group_by(programName,germplasmName) %>%
slice(1) %>% ungroup()
print(paste0(nrow(germNamesWithGenos_NoLocalSourcedGBS)," germNames without local GBS genos"))
genosForPhenos<-bind_rows(germNamesWithGenos_HasLocalSourcedGBS,
germNamesWithGenos_NoLocalSourcedGBS) %>%
bind_rows(germNamesWithDArT)Check for new germplasm “TMS20” and match DArT names to germplasmName.
genosForPhenos %<>%
bind_rows(readxl::read_xls(here::here("data/DatabaseDownload_2021Aug08",
"8885_genotyping_plate_layouts.xls")) %>%
select(`Sample ID`) %>%
separate(`Sample ID`,c("FullSampleName","germplasmName"),"\\|\\|\\|") %>%
mutate(FullSampleName=gsub("\\.","_",FullSampleName),
OrigKeyFile="DArTseqLD",
OriginOfSample="IITA",
programName="IITA"))print(paste0(nrow(genosForPhenos)," total germNames with genos either GBS or DArT"))dbdata %<>%
left_join(genosForPhenos)
# Create a new identifier, GID
## Equals the value SNP data name (FullSampleName)
## else germplasmName if no SNP data
dbdata %<>%
mutate(GID=ifelse(is.na(FullSampleName),germplasmName,FullSampleName))WARNING: User input required! If I had preselected locations before downloading, this wouldn’t have been necessary.
Based on previous locations used for IITA analysis, but adding based on list of locations used in IYR’s trial list data/2019_GS_PhenoUpload.csv: “Ago-Owu” wasn’t used last year.
dbdata %<>%
filter(locationName %in% c("Abuja","Ago-Owu","Ibadan","Ikenne","Ilorin","Jos","Kano",
"Malam Madori","Mokwa","Ubiaja","Umudike","Warri","Zaria"))
nrow(dbdata) # [1] 497414saveRDS(dbdata,file=here::here("output","IITA_CleanedTrialData_2021Aug08.rds"))The next step is to check the experimental design of each trial. If you are absolutely certain of the usage of the design variables in your dataset, you might not need this step.
Examples of reasons to do the step below:
One reason it might be important to get this right is that the variance among complete blocks might not be the same among incomplete blocks. If we treat a mixture of complete and incomplete blocks as part of the same random-effect (replicated-within-trial), we assume they have the same variance.
Also error variances might be heterogeneous among different trial-types (blocking scheme available) and/or plot sizes (maxNOHAV).
Start with cleaned data from previous step.
rm(list=ls()); gc()
library(tidyverse); library(magrittr);
source(here::here("code","gsFunctions.R"))
dbdata<-readRDS(here::here("output","IITA_CleanedTrialData_2021Aug08.rds"))dbdata %>% head %>% rmarkdown::paged_table()Detect designs
dbdata<-detectExptDesigns(dbdata)dbdata %>%
count(programName,CompleteBlocks,IncompleteBlocks) %>% rmarkdown::paged_table()saveRDS(dbdata,file=here::here("output","IITA_ExptDesignsDetected_2021Aug08.rds"))
sessionInfo()