ghap.predictblup {GHap}

Predict BLUP from reference

Description

Prediction of BLUP values in test individuals based on reference individuals.

Usage

  ghap.predictblup(refblup, vcp, covmat,
                   errormat = NULL,
                   errorname = "",
                   include.ref = TRUE,
                   diagonals = FALSE,
                   tol = 1e-12)

Arguments

refblup

A named numeric vector of reference BLUP values.

vcp

A numeric value for the variance in BLUP values.

covmat

A square matrix containing correlations among individuals. Both test and reference indiviudals must be present in the matrix.

errormat

A square error matrix for reference individuals. This matrix can be obtained with argument extras = "LHSi" in the ghap.lmm function.

errorname

The name used for the random effect in the ghap.lmm function. If the error matrix was imported from somewhere else, this argument can be ignored provided that the names in the error matrix match the ones in the covariance matrix.

include.ref

A logical value indicating if reference individuals should be included in the output (default = TRUE).

diagonals

A logical value indicating if diagonals of the covariance matrix should be used in calculations of accuracy and standard errors (default = FALSE). The default is to set diagonals to 1. For genomic estimated breeding values, using TRUE will account for inbreeding in the computation of accuracies and standard errors.

tol

A numeric value specifying the scalar to add to the diagonal of the covariance matrix if it is not inversible (default = 1e-12).

Value

A data frame with predictions of BLUP values. If an error matrix is provided, standard errors and accuracies are also included.

Author(s)

Yuri Tani Utsunomiya <[e-mail]>

References

J.F. Taylor. Implementation and accuracy of genomic selection. Aquaculture 2014. 420, S8-S14.

Examples

# #### DO NOT RUN IF NOT NECESSARY ###
# 
# # Copy plink data in the current working directory
# exfiles <- ghap.makefile(dataset = "example",
#                          format = "plink",
#                          verbose = TRUE)
# file.copy(from = exfiles, to = "./")
# 
# # Copy metadata in the current working directory
# exfiles <- ghap.makefile(dataset = "example",
#                          format = "meta",
#                          verbose = TRUE)
# file.copy(from = exfiles, to = "./")
# 
# # Load plink data
# plink <- ghap.loadplink("example")
# 
# # Load phenotype and pedigree data
# df <- read.table(file = "example.phenotypes", header=T)
# 
# ### RUN ###
# 
# # Subset individuals from the pure1 population
# pure1 <- plink$id[which(plink$pop == "Pure1")]
# plink <- ghap.subset(object = plink, ids = pure1, variants = plink$marker)
# 
# # Subset markers with MAF > 0.05
# freq <- ghap.freq(plink)
# mkr <- names(freq)[which(freq > 0.05)]
# plink <- ghap.subset(object = plink, ids = pure1, variants = mkr)
# 
# # Compute genomic relationship matrix
# # Induce sparsity to help with matrix inversion
# K <- ghap.kinship(plink, sparsity = 0.01)
# 
# # Fit mixed model
# df$rep <- df$id
# model <- ghap.lmm(formula = pheno ~ 1 + (1|id) + (1|rep),
#                   data = df,
#                   covmat = list(id = K, rep = NULL),
#                   extras = "LHSi")
# refblup <- model$random$id$Estimate
# names(refblup) <- rownames(model$random$id)
# 
# # Predict blup of reference and test individuals
# blup <- ghap.predictblup(refblup, vcp = model$vcp$Estimate[1],
#                          covmat = as.matrix(K),
#                          errormat = model$extras$LHSi,
#                          errorname = "id")
# 
# # Compare predictions
# plot(blup$Estimate, model$random$id$Estimate)
# abline(0,1)