# Genomic Best Linear Unbiased Predictors Analysis¶

## Performing GBLUP Analysis¶

The GBLUP method computes or imports a genomic relationship matrix and from that computes the “Genomic Best Linear Unbiased Predictor” (GBLUP) of additive genetic merits by sample and of allele substitution effects (ASE) by marker. [VanRaden2008], [Taylor2013]

Note

This method uses (with a genotypic spreadsheet) or assumes (with a numerically recoded spreadsheet) an additive genetic model.

## Large N Considerations¶

If your dataset consists of more than 8,000 samples (5,500 for 32-bit systems), you will first be presented with the following dialog:

Please see *Summary of Performance Tradeoffs* to know how best to respond to this
prompt. Afterward, you will be taken to the standard GBLUP options
explained below.

Note

As noted in this dialog, please select *Exact* if you plan
to use the AI REML algorithm.

## Options¶

**REML Computation Algorithm:**Choose the variant of the Restricted Maximum Likelihood method for estimating the genetic variance explained by the genotypes:*Use EMMA*: Usually faster. Required for predicting missing phenotypes.*Use AI REML*: The Average Information algorithm allows integrating multiple GRMs into the model, either for using an additional GRM for the gender chromosome, or for correcting for Gene by Environment interactions. See [Yang2011].

**Impute missing genotypic data as:**Missing genotypic data can be imputed by either of the following methods:*Homozygous major allele*: All missing genotypic data will be recoded to 0.*Numerically as average value*: All missing genotypic data will be recoded to the average of all non-missing genotype calls (using the additive model).Note

If

**Correct for Gender**(see below) is also selected, and there is non-missing data for both males and females in a given marker, averages for males and females will be computed and used separately.

**Correct for Gender**: Assumes the column is coded as if the male were homozygous for the X-Chromosome allele in question. Uses the [Taylor2013] gender-correction algorithm. (See*Correcting the GRM for Gender Using Overall Normalization*and*Correcting the GRM for Gender Using Normalization by Individual Marker*.) Two values of the ASE are output, one for each gender.**Choose Sex Column**: Choose the spreadsheet column that specifies the gender of the sample. This column may either be categorical (“M” vs. “F”) or binary (0 = male, 1 = female).**Chromosome that is hemizygous for males**: Usually the X Chromosome, which is the default.**Dosage compensation**: Modify the dosage compensation.**GRM for Gender Chromosomes**: The AI REML allows for multiple GRMs to be used in a single analysis. You can compute a GRM for your gender chromosomes separately and specify the precomputed GRM here. To be valid, this spreadsheet must follow the rules outlined in*Precomputed Kinship Matrix Option*.

**Use Pre-Computed Genomic Relationship Matrix**: To use, check this option, then click on**Select Sheet**and select the genomic relationship matrix spreadsheet from the window that is presented. To be valid, this spreadsheet must follow the rules outlined in*Precomputed Kinship Matrix Option*.Note

When using a pre-computed genomic relationship matrix, the matrix and the scaling factor are re-calculated from the genotypic data being used for this analysis.

**Normalization Algorithm (Used or Assumed) for the GRM:**If a pre-computed GRM is not selected, this choice influences the GRM computation. Otherwise, this choice will still influence ASE computations. See [Yang2011] for details of the individual marker normalization. Overall normalization is the default. (See*The Genomic Relationship Matrix*.)**Correct for Additional Covariates**: Allows additional fixed effects to be added to this model from columns of this spreadsheet. Fixed effect covariates can be binary, integer, real-valued, categorical or (if actual genotypic data rather than recoded genotypic data is being used for the analysis) genotypic. In all cases, if a marker is used as an additional fixed effect, it will not be included in the analysis in any other way. To begin, check this option, then click on**Add Columns**to get a choice of spreadsheet columns to use.**Correct for Gene by Environment Interactions**: Allows for the correct for gene by environment interactions based on an environment variable, which may be binary or categorical. (See*Incorporating Gene by Environment Interactions into GBLUP*.) To begin, check this option, then click on**Add Columns**to get a choice of spreadsheet columns to use.Note

You must use the AI REML algorithm to correct for gene by environment interactions.

Note

Ouputs for gene by environment interactions will always be as described in

*Spreadsheet Output for Multiple Random Effects*and in*Node Change Log Output for the AI Algorithm*.**Missing Phenotypes**: To predict random effects (genomic merit/genomic breeding values) and the phenotypes for samples with missing phenotypes, select**Predict random effects for samples with missing phenotypes**. Selecting this will also include samples with missing phenotypes as a part of the basis for the ASE calculations. Otherwise, select**Drop samples with missing phenotypes**.Note

You must use the EMMA algorithm to predict random effects.

## GBLUP Output¶

### Spreadsheet Output for One Random Effect¶

The following two spreadsheets will always be created:

GBLUP estimates by sample: This spreadsheet contains a column for the phenotype (selected dependent variable) of each sample and a column for theRandom effects componentfor each sample.

GBLUP estimates by marker: These are the GBLUP estimates of the allele substitution effect (ASE) by marker, along with the absolute magnitude of the ASE and, if EMMA was used, the normalized absolute magnitude of the ASE. If gender correction is applied, separate columns for the ASE, the absolute magnitude of the ASE, and (with EMMA) the normalized Abs ASE will be output for both males and females.The marker map will be applied to this spreadsheet.

### Spreadsheet Output for Multiple Random Effects¶

The following two spreadsheets will always be created:

GBLUP estimates by sample: This spreadsheet contains a column for the phenotype (selected dependent variable) of each sample and, for each random effect, a column containing theRandom effects componentrelated to that random effect for each sample.

GBLUP estimates by marker: For each random effect, this spreadsheet contains a column for the GBLUP estimates of the allele substitution effect (ASE) by marker resulting from this random effect and a column for the absolute magnitude of the ASE from this random effect. If gender correction is applied, separate columns for the ASE and the absolute magnitude of the ASE will be output for both males and females for each random effect.The marker map will be applied to this spreadsheet.

### Node Change Log Output for the EMMA Algorithm¶

If the EMMA algorithm is used, the following will be output to the node change log of each spreadsheet (in addition to the options used and the summary statistics of numbers of records processed, etc.):

The pseudo-heritability , which is

The would-be pseudo-heritability calculated as if the scaled genomic relationship matrix had been used. This is

where is the scaling factor that would have been necessary to scale the genomic relationship matrix according to the methodology of

*Scaling the Kinship Matrix*.The would-be scaling factor .

The variance and standard error of the pseudo-heritability and , see

*Estimating the Variance of Heritability for One Random Effect*for the formula.The p-value, which is

where is chi-square distributed with one degree of freedom, is the restricted maximum-likelihood (REML) estimate (

*Finding the Variance Components Using EMMA*), and is the restricted log-likelihood for the corresponding linear model with no random effects (). may be written asNote

The above expression for may be obtained by taking the expression for found in

*Finding the Variance Components Using EMMA*and allowing to tend towards infinity, so that tends towardswhich is independent of .

The genetic component of variance

*Vg*().The error component of variance

*Ve*().

If you have selected additional covariates, two additional outputs will appear in these node change logs:

*Proportion of genetic variance*and*Prop. explained by fixed covariates*.

These correspond exactly to the proportion of *Genetic variance*
( for ) and the proportion of *Variance
explained* () outputs of the **Variance Partition
Plot** (See *The Variance Partition Plot*) for the initial model
() of an MLMM run. (The pseudo-heritability
for the unnormalized genomic relationship matrix is
used here.)

- Any messages from the algorithm concerning steps that had to be taken to make the algorithm converge.

### Node Change Log Output for the AI Algorithm¶

If the Average Information (AI) algorithm is used, the following will be output to the node change log of each spreadsheet (in addition to the options used and the summary statistics of numbers of records processed, etc.):

The variances for the full model.

First, the number of iterations required for convergence and the log(likelihood) for this model are output.

Then, a table is output containing columns for the description of (

*Source*), the value of (*Variance*), and the standard error (*SE*) of- Each variance component (including
*V(e)*()) *Vp*(), the sum of all variance components (including*V(e)*())- Each random-effect variance component divided by
*Sum of V(G)/Vp*, the sum of all random-effect variance components divided by , will be shown if there is more than one random effect being analyzed.

- Each variance component (including
If there is more than one random effect being analyzed, the variances for each of:

- the full model without the first random effect
- the full model without the second random effect
- ...
- the full model without the last random effect.

The same output format is used as for the variances for the full model.

The variances for “the completely reduced model”–that is, the model containing only the error term .

The same output format is used as for the variances for the full model.

If there is more than one random effect being analyzed, a likelihood ratio test for each of:

- the full model vs. the full model without the first random effect
- the full model vs. the full model without the second random effect
- ...
- the full model vs. the full model without the last random effect.

The following is output for each random effect:

*logL*The full-model likelihood*logL0*The likelihood of the full model without this random effect*LRT*The likelihood ratio test ()*df*The degrees of freedom for this test*pval*The p-value for this test

An overall random effects likelihood test. The following is output for this test:

*logL*The full-model likelihood*logL0*The likelihood of the null model*LRT*The likelihood ratio test ()*df*The degrees of freedom for this test*pval*The p-value for this test

Sampling variance/covariance of the estimates of variance components for the full model.

A square table is output with a row and a column for each of the variance components. The diagonal contains the sampling variance of each variance component, while each off-diagonal element contains the covariance between the row’s variance component and the column’s variance component.

Any messages from the algorithm concerning steps that had to be taken to make the algorithm converge.

## GBLUP Genomic Relationship Matrix¶

Unless the **Use Pre-Computed Genomic Relationship Matrix** option is
selected, a **GBLUP Genomic Relationship Matrix** spreadsheet will be
created. This spreadsheet can be used not only as a pre-computed
relationship matrix for other runs of this GBLUP tool, but also as a
pre-computed kinship matrix for the EMMAX and MLMM Mixed Model GWAS
methods (*Mixed Linear Model Analysis*).

This matrix can also be computed when there is no dependent variable
available. See *Separately Computing the Genomic Relationship Matrix* for more information.