BLUP/BLUE Calculator

Owned by Star Yanxin Gao
Last updated: May 12, 2020
4 min read

Created by Star Yanxin Gao, last modified by @Umesh Rosyara on Feb 27, 2019

Best linear unbiased prediction (BLUP) is estimate of random effects and Best linear unbiased estimator (BLUE) is estimate of fixed effects in linear mixed models. This tool calculate BLUP or BLUE associated with genotype / variety. 

What it does

This function calculates BLUP or BLUE for Y variables based on the model specified by the user. The BLUE/BLUPs are calculated in batch for variables between start and stop variables. So, the Y variables should be arranged together for this purpose.

This version supports to two Designs – Random Complete Block Design (RCBD) and Latin Square Design (LSD). The required variables for RCBD are Genotype and Replication and for lattice required variables are Genotype, Replication (main block) and Block within Replication (sub block or block).

User can add additional up to two X variables (eg. Location, Season etc) in the model and are treated as factor (even supplied as numerical variable). If you want to calculate BLUP / BLUE by a grouping variable (eg. Management), then this will calculate BLUP / BLUE for each level of the variable.

BLUP/BLUE calculator (multiple traits) (Galaxy Version 1.11.0)

Step1: Get data:

(required) Must be the TAB file as both column delimiter and file type (you can save excel files as “tab delimited text” file with first row has column header names)

Example test dataset:

Phenotype_BLUP_BLUE-demo.tab

Step 2: Choose Design 

(required) choose available design options (Lattice or RCBD) from dropdown menu

Step 3: Select the appropriate name of columns used as input for Block with Replication ( for lattice design only), Replication column, Genotype column and Y column. 

Replication Column

(required) select replication variable column name

Genotype Column

(required) Must be the genotype variable column name

Genotype here indicates variety/lines / accessions or any other primary variables whose Best Linear Unbiased Estimates (BLUE) and best linear unbiased prediction (BLUPs) need to be calculated.

Start Y Column

(required) select names of first Y variable name from where BLUP/BLUE calculation starts.

 

End Y Column

(required) select names of last Y variable name from where BLUP/BLUE calculation ends. The BLUE/BLUP will be calculated between start and end Y column. If start and end column are same, BLUP / BLUE will be calculated for only one selected variable.

Summarize by Column

(optional) if you want to calculate BLUP/BLUE for every level for summary variable (for example for each year separately then Year should be “summarize by variable”)

 

The basic model (without additional variables).

Y = Genotype + Replication + error RCBD

Y = Genotype + Replication + Block within Replication + error  Lattice 

Step 4: Summarized by

Summarized by loop the above model for each level of summarized by variable. For example if we are summarized by variable is Management, this will fit the above model for each level of the variable.

for ( i in 1: number of management){

     Y = Genotype + Replication + Block within Replication + error 

}  

Step 6: Additional variables 

The tool can fit maximum two additional variables in model variable 1 (factor) and variable 2 (factor). This optional to tool. The model implemented by choosing factor 1 and factor 2 is as follows:

Include additional Variable 1

(optional) choose additional first variable (factor) – see below description

 Name of Variable 1

Select name of group variable name (required if include additional variable 1 is Yes)

 Include additional Variable 2

(optional) choose additional second variable in addition to variable 1 (factor)

 Variable 2, Factor Column

It can choose any column as the group variable 2 (factor)

 

 

Report BLUE or BLUP in summary

Choose whether you want to calculate BLUE or BLUP

 

 Model

for RCBD design

without additional factors (variable 1 and variable 2)

Y = Genotype + Replication + error RCBD

Y = Genotype + Replication + Block within Replication + error Lattice

With variable 1

Y = Variable 1 + Genotype + Variable 1 : Genotype + Replication + error RCBD

Y = Variable 1 + Genotype + Replication + Variable 1 : Genotype + Block within Replication + error Lattice

With variable 1 and variable 2

Y = Variable 2 + Variable 1 + Genotype + Variable 1 : Genotype + Variable 1: Variable 2 + Variable 1:Variable 2: Genotype + Replication + error RCBD Y = Variable 2 + Variable 1 + Genotype + Variable 1 : Genotype + Variable 1: Variable 2 + Variable 1:Variable 2: Genotype + Replication + Block within Replication + error RCBD

Model details on the model

  • Genotype is treated as Fixed for BLUE calculation and random for BLUP calculation.

  • Replication, Block within Replication are treated as random both BLUP and BLUE calculations.

  • Variable 1, Genotype: Variable 1 are treated as random both BLUP and BLUE calculations.

  • Variable 2, Genotype: Variable 2, Variable 1: Variable 2, Genotype: Variable 1 : Variable 2 are treated as random effects for both BLUP and BLUE calculations.

 

Backend Source Code

This version of BLUP/BLUE calculator is based on lme4 R package version 1.1_18_1.

 

Contributors:

The methodology and R script contributors include Umesh Rosyara (CIMMYT), Juan David Arbelaez Velez (IRRI) and Juan Burgueno (CIMMYT). The Galaxy integration contributor Angel Villahoz-Baleta (Cornell University) and Dereeper Alexis.

 

References

Douglas Bates, Martin Maechler, Ben Bolker, Steve Walker (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1-48.

 

Corresponding contract:

 For any questions about this tool, please send an e-mail to u.rosyara@cgiar.org

 

Acknowledgment:

This tool is developed based on GOBii project and implemented in EiB server