,
Humberto Fanelli Carvalho [aut]
, Julio Isidro y
Sanchez [aut] | Title: | Best Linear Unbiased Prediction of Genotype-by-Environment Metrics |
|---|---|
| Description: | Provides functions to calculate the best linear unbiased prediction of genotype-by-environment metrics: ecovalence, environmental variance, Finlay and Wilkinson regression and Lin and Binns superiority measure, based on a multi-environment genomic prediction model. |
| Authors: | Simon Rio [aut, cre] ,
Humberto Fanelli Carvalho [aut]
, Julio Isidro y
Sanchez [aut] |
| Maintainer: | Simon Rio <[email protected]> |
| License: | GPL (>= 3) |
| Version: | 1.0.0 |
| Built: | 2025-03-02 03:02:39 UTC |
| Source: | https://github.com/cran/GEmetrics |
This function formats the phenotypic data as well as the kinship matrix for BGLR
BGLR_format(Pheno, K)BGLR_format(Pheno, K)
Pheno |
a data frame with three columns: "Y" for phenotypic values, "Genotype" for genotype names and "Environment" for environment names. All genotypes names must be included in the set or row/column names of the "K" matrix. The number of environments (J) must be at least two |
K |
a square kinship or genomic relationship matrix for N genotypes whose row and column names include those of the "Genotype" column of the "Pheno" matrix |
a list of two elements: a "BGLR_pheno" phenotypic response matrix with J columns to be used in BGLR and the corresponding "BGLR_K" kinship matrix.
## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 75% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*3/4)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Generate the phenotypic response matrix for BGLR and the corresponding kinship matrix BGLR_data <- BGLR_format(Pheno=DataSim$Pheno,K=wheat.A) head(BGLR_data$BGLR_pheno)## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 75% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*3/4)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Generate the phenotypic response matrix for BGLR and the corresponding kinship matrix BGLR_data <- BGLR_format(Pheno=DataSim$Pheno,K=wheat.A) head(BGLR_data$BGLR_pheno)
This function calculates the best linear unbiased prediction and the conditional variance matrix of environment-specific breeding values
EnvBV_blup(Pheno, K, Omega_G, Omega_E)EnvBV_blup(Pheno, K, Omega_G, Omega_E)
Pheno |
a data frame with three columns: "Y" for phenotypic values, "Genotype" for genotype names and "Environment" for environment names. All genotypes names must be included in the set or row and column names of the "K" matrix. The number of environments must be at least two |
K |
a square kinship or genomic relationship matrix for N genotypes whose row and column names include those of the "Genotype" column of the "Pheno" matrix |
Omega_G |
a square matrix with genetic covariances between J environments whose row and column names correspond to those of the "Environment" column of the "Pheno" matrix |
Omega_E |
a square matrix with error covariances between J environments whose row and column names correspond to those of the "Environment" column of the "Pheno" matrix |
a list of two elements: a "G_hat" matrix of best linear unbiased prediction of environment-specific breeding values for the N genotypes (as rows) in J environments (as columns), and a square conditional variance matrix "P" of environment-specific breeding values of dimension NJxNJ where each row block of size N correspond to an environment and the rows of each block correspond to genotypes
## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 80% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*4/5)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Calculate the blup and the conditional variance matrix using simulated variance components ## this step can take several seconds ## note that variance can also be estimated (e.g. using BGLR) BlupEnvBV <- EnvBV_blup(Pheno=DataSim$Pheno,K=wheat.A,Omega_G=DataSim$Omega_G, Omega_E=DataSim$Omega_E) ## Display results head(BlupEnvBV$G_hat) BlupEnvBV$P[1:5,1:5]## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 80% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*4/5)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Calculate the blup and the conditional variance matrix using simulated variance components ## this step can take several seconds ## note that variance can also be estimated (e.g. using BGLR) BlupEnvBV <- EnvBV_blup(Pheno=DataSim$Pheno,K=wheat.A,Omega_G=DataSim$Omega_G, Omega_E=DataSim$Omega_E) ## Display results head(BlupEnvBV$G_hat) BlupEnvBV$P[1:5,1:5]
This function calculates the best linear unbiased prediction of the following GE metrics: ecovalence, environmental variance, Finlay and Wilkinson regression and Lin and Binns superiority measure. Ignoring the P matrix resumes to ignoring the condition variance term in the calculation
GEmetrics_blup(G_hat, metric, P = NULL)GEmetrics_blup(G_hat, metric, P = NULL)
G_hat |
a matrix of best linear unbiased prediction of environment-specific breeding values for N genotypes (as rows) in J environments (as columns) |
metric |
a character string indicating what GE metric to consider: "Ecovalence", "EnvironmentalVar", "FinlayWilkRegression", "LinBinns" |
P |
(optional) a square conditional variance matrix of environment-specific breeding values of dimension NJxNJ, where each row block of size N correspond to an environment and the rows of each block correspond to genotypes |
a vector of size N with best linear unbiased prediction of the GE metric
## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 80% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*4/5)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Calculate the blup and the conditional variance matrix using simulated variance components ## this step can take several seconds ## note that variance can also be estimated (e.g. using BGLR) BlupEnvBV <- EnvBV_blup(Pheno=DataSim$Pheno,K=wheat.A,Omega_G=DataSim$Omega_G, Omega_E=DataSim$Omega_E) ## Compute GE metric estimates GEmetrics_hat <- GEmetrics_blup(G_hat=BlupEnvBV$G_hat,metric="Ecovalence",P=BlupEnvBV$P) head(GEmetrics_hat)## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 80% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*4/5)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Calculate the blup and the conditional variance matrix using simulated variance components ## this step can take several seconds ## note that variance can also be estimated (e.g. using BGLR) BlupEnvBV <- EnvBV_blup(Pheno=DataSim$Pheno,K=wheat.A,Omega_G=DataSim$Omega_G, Omega_E=DataSim$Omega_E) ## Compute GE metric estimates GEmetrics_hat <- GEmetrics_blup(G_hat=BlupEnvBV$G_hat,metric="Ecovalence",P=BlupEnvBV$P) head(GEmetrics_hat)
This function calculates the best linear unbiased prediction and the conditional variance matrix of environment-specific breeding values
Simulate_MET_data(Design, K, h2 = 0.5, rho = 0.5, sd_mu = 1)Simulate_MET_data(Design, K, h2 = 0.5, rho = 0.5, sd_mu = 1)
Design |
a data frame with two columns: "Genotype" for genotype names and "Environment" for environment names. All genotypes names must be included in the set of row/column names of the "K" matrix. The number of environments (J) must be at least two |
K |
a square kinship or genomic relationship matrix for N genotypes whose row/column names include those of the "Genotype" column of the "Design" matrix |
h2 |
heritability (numeric value between 0 and 1 excluded) of observations in each environment: either a scalar to set a common heritability for all environments, or a vector of heritabilities associated with each environment of size J. The default value is a heritability of 0.5 for all environments |
rho |
genetic correlation (numeric value between -1 and 1 excluded) between environment pairs: either a scalar to set a common genetic correlation between all environment pairs, or a square correlation matrix of dimension JxJ. The default value is a genetic correlation of 0.5 between all environment pairs |
sd_mu |
standard deviation (positive numeric value) of the Gaussian distribution in which environment means are drawn. The default value is 1 |
a list of two elements: a "Pheno" data frame consisting of the "Design" data frame to which a "Y" column containing simulated phenotypic values has been added, and a "EnvBV" matrix of dimension NxJ containing the simulated environment-specific breeding values
## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 75% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*3/4)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Simulated phenotypes head(DataSim$Pheno) ## Simulated environment-specific breeding values head(DataSim$EnvBV) ## Genetic covariance matrix between environments DataSim$Omega_G ## Error covariance matrix between environments DataSim$Omega_E## Set seed for reproductibility set.seed(123) ## Load "wheat" dataset from BGLR data("wheat",package = "BGLR") ## Generate a design data frame for all genotypes in 5 environments Design <- expand.grid(Genotype=rownames(wheat.A),Environment=paste0("Env",1:5)) ## Set sparseness by discarding 75% of the combinations Design <- Design[-sample(nrow(Design),round(nrow(Design)*3/4)),] ## Simulate phenotypic data with default parameter values DataSim <- Simulate_MET_data(Design=Design,K=wheat.A) ## Simulated phenotypes head(DataSim$Pheno) ## Simulated environment-specific breeding values head(DataSim$EnvBV) ## Genetic covariance matrix between environments DataSim$Omega_G ## Error covariance matrix between environments DataSim$Omega_E