italo-granato / bgge

Package to genomic prediction focused on GE genomic models

Bayesian Genomic Linear Models Applied to GE Genome Selection - Development version 0.6.5

[Last README update: 2018-11-13]

Table of contents

• NEWS
• Instructions
  • Installation
  • Load the package
  • Quick use
  • Other Params
• How to cite this package
• Contributions
• Authors

News of this version (0.6.5)

Changes:

• Added compatibility with the functions (print() and plot()).
• Include ORCID of the authors.
• Include url of the project.
• Now the BGGE() function exports the input parameters.

See the last updates in NEWS.

Instructions for proper implementation

Installation

To complete installation of dev version of BGGE from GitHub, you must have previously installed the devtools package.

install.packages('devtools')
devtools::install_github('italo/BGGE')

If you want to use the stable version of BGGE package, install it from CRAN.

install.packages('BGGE')

Load the package

library(BGGE)

Example of simple usage of the package

library(BGGE)
library(BGLR)
data(wheat)

X <- scale(wheat.X, scale = TRUE, center = TRUE)
rownames(X) <- 1:599
pheno_geno <- data.frame(env = gl(n = 4, k = 599), 
                         GID = gl(n=599, k=1, length = 599*4),
                         value = as.vector(wheat.Y))

# Creating kernel for GE model

K <- getK(Y = pheno_geno, X = X, kernel = "GB", model = "MM")
y <- as.vector(wheat.Y)

fit <- BGGE(y = y, K = K, ne = rep(599, 4))
fit

## Model Fitted with: 
##  1000  Iterations, burning the first  200  and thining every  3 
## 
##  Some predicted Values: 
##  [1]  -0.2737  -0.4081  -0.3799   0.3694  -0.0362   0.2354   0.2840
##  [8]  -0.1194  -0.7742  -1.0280
## 
##  Use str() function to found more datailed information.

plot(fit)

Others params

params	Use
XF	Design matrix for fixed effects.
ite	Number of iterations.
ne	Number of subjects by environment.
burn	Number of iterations to be discarded as burn-in.
thin	Thinin interval.
verbose	Should report be printed on screen?
tol	tolerance for zero. Default is 1e-10
R2	the proportion of variance expected to be explained by the regression

Citation

First option, by the paper.

@article {Granato2018,
    author = {Granato, Italo and Cuevas, Jaime and Luna-V{\'a}zquez, Francisco J. and Crossa, Jos{\'e} and Montesinos-L{\'o}pez, Osval A. and Burgue{\~n}o, Juan and Fritsche-Neto, Roberto},
    title = {BGGE: A New Package for Genomic-Enabled Prediction Incorporating Genotype {\texttimes} Environment Interaction Models},
    year = {2018},
    doi = {10.1534/g3.118.200435},
    publisher = {G3: Genes, Genomes, Genetics},
    URL = {http://www.g3journal.org/content/early/2018/07/25/g3.118.200435},
    eprint = {http://www.g3journal.org/content/early/2018/07/25/g3.118.200435.full.pdf},
    journal = {G3: Genes, Genomes, Genetics}
}

Second option, by the package

citation('BGGE')

## 
## To cite package 'BGGE' in publications use:
## 
##   Italo Granato, Luna-Vázquez Francisco J. and Cuevas Jaime
##   (2018). BGGE: Bayesian Genomic Linear Models Applied to GE
##   Genome Selection. R package version 0.6.5.
##   https://github.com/italo-granato/BGGE
## 
## A BibTeX entry for LaTeX users is
## 
##   @Manual{,
##     title = {BGGE: Bayesian Genomic Linear Models Applied to GE Genome Selection},
##     author = {Italo Granato and Luna-Vázquez {Francisco J.} and Cuevas Jaime},
##     year = {2018},
##     note = {R package version 0.6.5},
##     url = {https://github.com/italo-granato/BGGE},
##   }

Contributions

If you have any suggestions or feedback, I would love to hear about it. Feel free to report new issues in this link, also if you want to request a feature/report a bug, or make a pull request if you can contribute.

Research and implementations

• BGGE: A New Package for Genomic-Enabled Prediction Incorporating Genotype × Environment Interaction Models

Authors

• Italo Granato (Author, Maintainer)
• Jaime D. Cuevas D. (Author)
• Francisco J. Luna-Vázquez (Author, Maintainer)

bgge's Issues

Confusion about variance components

Hi,

I'm running BGGE using the maizefiles.RData and copying the code in box 5 of the BGGE paper. The only difference from box 5 is i'm using a GBLUP kernel ie:

library(BGGE)

### Load the maize dataset from supplementary material

load(“maizefiles.Rdata”)

ne <- as.vector(table(pheno_geno$env))

K2 <- getK(Y = pheno_geno, X=geno, kernel = “GB”, bandwidth = 1, model = “MDe”)

fit <- BGGE(y = pheno_geno$GY, K = K2, ne = ne)

fit$yHat[pheno_geno$env == “AN_LN”] #predicted values for environment 2

fit$K$G$varu #main genetic variance

fit$varE #residual variance

fit$K$AN_LN$varu #specific genetic variance

fit$varE #residual variance

plot(fit$yHat, pheno_geno$GY)

What I don't understand is how to reconcile these results with the variance components listed in table 2 of the BGGE paper. The code above gives me values around 1.8 for fit$varE and around 3.3 for fit$K$G$varu, but table 2 seems to suggest there should be more residual variance than variance explained by genetic effects. Is the data used to make table 2 in the BGGE paper the same as maizefiles.Rdata? How can I properly extract the variance components from the fit like was done to produce table 2?

Thanks!