R package:blupADC-Feature 6

Last updated on Apr 20, 2021 6 min read DMU

Table of Contents

Overview

🤠 In the previous section, we have given detailed description about data preparation. In the following section, we will introduce genetic evaluation software in animal and plant breeding. Nowadays, in the filed of animal and plant breeding, two of the most famous breeding software are DMU and BLUPF90 ( cited over than one thousand).

Although these two softwares have many advantages, these two softwares have one common pitfall: it is a little difficult to use for freshman(need to prepare parameter file). Thus, in order to overcome this pitfall, package blupADC provides run_DMU and run_BLUPF90 for interfacing DMU and BLUPF90 in an easy way.

In this section, we will give detail descriptions about run_DMU function.

👉 Note: Package blupADC has encapsulated the basic module of DMU(dmu1,dmuai, and dmu5),more modules could be download from website(DMU download website).

For commercial use of DMU, user must contact the author of DMU !!!

Example

Single trait model - Pedigree BLUP

library(blupADC)
data_path=system.file("extdata", package = "blupADC")  #  path of provided files 
  
run_DMU(
        phe_col_names=c("Id","Mean","Sex","Herd_Year_Season","Litter","Trait1","Trait2","Age"), # colnames of phenotype
        target_trait_name=c("Trait1"),                           #trait name 
        fixed_effect_name=list(c("Sex","Herd_Year_Season")),     #fixed effect name
        random_effect_name=list(c("Id","Litter")),               #random effect name
        covariate_effect_name=NULL,                              #covariate effect name
        phe_path=data_path,                          #path of phenotype file
        phe_name="phenotype.txt",                    #name of phenotype file
        integer_n=5,                                 #number of integer variable 
        analysis_model="PBLUP_A",                    #model of genetic evaluation
        dmu_module="dmuai",                          #modeule of estimating variance components 
        relationship_path=data_path,                 #path of relationship file 
        relationship_name="pedigree.txt",            #name of relationship file 
        output_result_path="/root"                   # output path 
        )

Single trait model - GBLUP

library(blupADC)
data_path=system.file("extdata", package = "blupADC")  #  path of provided files 
run_DMU(
        phe_col_names=c("Id","Mean","Sex","Herd_Year_Season","Litter","Trait1","Trait2","Age"), # colnames of phenotype 
        target_trait_name=c("Trait1"),                           #trait name 
        fixed_effect_name=list(c("Sex","Herd_Year_Season")),     #fixed effect name
        random_effect_name=list(c("Id","Litter")),               #random effect name
        covariate_effect_name=NULL,                              #covariate effect name
        phe_path=data_path,                          #path of phenotype file
        phe_name="phenotype.txt",                    #name of phenotype file
        integer_n=5,                                 #number of integer variable 
        analysis_model="GBLUP_A",                    #model of genetic evaluation
        dmu_module="dmuai",                          #modeule of estimating variance components 
        relationship_path=data_path,                 #path of relationship file 
        relationship_name="G_Ainv_col_three.txt",            #name of relationship file 
        output_result_path="/root"                   # output path 
        )

Single trait model - Single-step BLUP

library(blupADC)
data_path=system.file("extdata", package = "blupADC")  #  path of provided files 
run_DMU(
        phe_col_names=c("Id","Mean","Sex","Herd_Year_Season","Litter","Trait1","Trait2"), # colnames of phenotype 
        target_trait_name=c("Trait1"),                           #trait name 
        fixed_effect_name=list(c("Sex","Herd_Year_Season")),     #fixed effect name
        random_effect_name=list(c("Id","Litter")),               #random effect name
        covariate_effect_name=NULL,                              #covariate effect name
        phe_path=data_path,                          #path of phenotype file
        phe_name="phenotype.txt",                    #name of phenotype file
        integer_n=5,                                 #number of integer variable 
        analysis_model="SSBLUP_A",                    #model of genetic evaluation
        dmu_module="dmuai",                          #modeule of estimating variance components 
        relationship_path=data_path,                 #path of relationship file 
        relationship_name=c("pedigree.txt","G_A_col_three.txt"),            #name of relationship file 
        output_result_path="/root"                   # output path 
        )

Through modifying the two parameters: analysis_model and relationship_name , we can perform Pedigree-BLUP, GBLUP, and SSBLUP analysis(PS: we can get G_Ainv_col_three.txt and G_A_col_three.txt through cal_kinship function，see more details about cal_kinship function).

The above example is single-trait model, while in actual breeding, multiple traits model is also common. Similarity, we only need to modify several parameters to perform multiple traits model:

Multiple traits model - Pedigree BLUP

library(blupADC)
data_path=system.file("extdata", package = "blupADC")  #  path of provided files 
  
run_DMU(
        phe_col_names=c("Id","Mean","Sex","Herd_Year_Season","Litter","Trait1","Trait2","Age"), # colnames of phenotype 
        target_trait_name=c("Trait1","Trait2"),                           #trait name 
        fixed_effect_name=list(c("Sex","Herd_Year_Season"),c("Herd_Year_Season")),     #fixed effect name
        random_effect_name=list(c("Id","Litter"),c("Id")),               #random effect name
        covariate_effect_name=list(NULL,"Age"),                              #covariate effect name
        phe_path=data_path,                          #path of phenotype file
        phe_name="phenotype.txt",                    #name of phenotype file
        integer_n=5,                                 #number of integer variable 
        analysis_model="PBLUP_A",                    #model of genetic evaluation
        dmu_module="dmuai",                          #modeule of estimating variance components 
        relationship_path=data_path,                 #path of relationship file 
        relationship_name="pedigree.txt",            #name of relationship file 
        output_result_path="/root"                   # output path 
        )

Parameter

🤡Basic

1：phe_path

File path of phenotype data ，character class。

2：phe_name

File name of phenotype data，character class。

Note: User-provided phenotype don’t have colnames (the same as the requirement of DMU)

3：phe_col_names

Column names of phenotype data，character class。

4：integer_n

Number of integer variable, numeric class。

5：genetic_effect_name

Genetic effect name(usually is the id name), character class.

6：target_trait_name

Target trait name, character class.

For multiple traits model, we should set target_trait_name as character vector, e.g. target_trait_name=c("Trait1","Trait2")

7：fixed_effect_name

Fixed effects name, list class.

For multiple traits model, the order of fixed effects name should corresponding to the target trait name.

eg. target_trait_name=c("Trait1","Trait2")

fixed_effect_name=list(c("Sex","Herd_Year_Season"),c("Herd_Year_Season"))

which means the fixed effects name of trait1 is : c("Sex","Herd_Year_Season"), the fixed effect name of trait2 is : c("Herd_Year_Season")

8：random_effect_name

Random effects name, list class.

For multiple traits model, the order of random effects name should corresponding to the target trait name.

eg. target_trait_name=c("Trait1","Trait2")

random_effect_name=list(c("Id","Litter"),c("Id"))

which means the random effects name of trait1 is : c("Id","Litter"), the random effects name of trait2 is : c("Id")

9：covariate_effect_name

Covariate effects name, list class.

For multiple traits model, the order of covariate effects name should corresponding to the target trait name.

eg. target_trait_name=c("Trait1","Trait2")

covariate_effect_name=list(NULL,"Age")

which means the covariate effects name of trait1 is : NULL (NULL means no this effect), the covariate effects name of trait2 is : Age

10：analysis_model

Model of genetic evaluation, character class.
- "PBLUP_A" : Pedigree BLUP- additive model
- "GBLUP_A" :GBLUP- additive model
- "GBLUP_AD" :GBLUP- additive and dominance model
- "SSBLUP_A" :SSBLUP- additive model
- "User_define": User define model
**11：dmu_module **

Module of estimating variance components, character class.
- "dmuai"
- "dmu4"
- "dmu5"
12：DMU_software_path

Path of DMU software, character class.

13：relationship_path

File path of relationship data, character class.

14：relationship_name

File name of relationship data, character class.

For difference genetic evaluation model , we should provide different relationship file.

E.g. for “PBLUP_A” model, we need to provide pedigree file, then we should set relationship_name="pedigree.txt" ;

for “GBLUP_A” model, we need to provide inverse of additive relationship matrix file(3 columns format), then we should set relationship_name="G_Ainv_col_three.txt" ;

for “SSBLUP_A” model, we need to provide pedigree and additive relationship matrix file(3 columns format), then we should set relationship_name=c("pedigree.txt","G_A_col_three.txt") ;

15：output_result_path

Path of output DMU result, character class.

16：output_ebv_path

File path of output EBV, character class. Default is equal to output_result_path

17：output_ebv_name

File name of output EBV, character class.

👺Advanced

18：provided_effect_file_path

File path of trait’s effect data, character class.

File of trait’s effect including fixed effects name, random effects name, and covariate effects name. Once user provide this file, user don’t need to set these three parameters: fixed_effect_name random_effect_name covariate_effect_name .

The format of this effect file is as following:

V1	V2	V3	V4	V5	V6	V7	V8	V9
Trait1	*	Sex	Herd_Year_Season	*	Id	Litter	*	*
Trait2	*	Sex	*	Id	*	Age	*

The first column is the name of target trait. Each column stands for one effect name. In order to recognize three types of effect, we set * to distinct each type.

Effects name between the first * and the second * stand for fixed effects name;

effects name between the second * and the third * stand for random effects name;

effects name between the third * and the fourth * stand for covariate effects name.

19：provided_effect_file_name

File name of trait’s effect data, character class.

20：provided_DIR_file_path

File path of user-provided DIR data, character class.

21：provided_DIR_file_name

File name of user-provided DIR data, character class.

22：included_permanent_effect

If perform permanent-environment analysis, logical class. Default is FALSE.

23：dmu_algorithm_code

Number of dmu-module algorithm, numeric class.

24：provided_prior_file_path

File path of user-provided prior file, character class.

25：provided_prior_file_name

File name of user-provided prior file, character class.

26：missing_value

Missing value in phenotype file, numeric class. Default is -9999.

27：iteration_criteria

Value of iteration convergence, numeric class. Default is 1.0e-7.

28：genetic_effect_number

Number of genetic effect in SOL file, numeric class. Default is 4.

29：residual_cov_trait

Traits combination of assuming residual-covariance equals to 0. e.g residual_cov_trait=list(c("Trait1","Trait2"))

30：selected_id

Individuals set of output EBV, character class.

31：cal_debv

If calculate desegregated-EBV(DEBV), logical class. Default is FALSE.

32：debv_pedigree_path

File path of pedigree data for calculating DEBV, character class.

33：debv_pedigree_name

File name of pedigree data for calculating DEBV, character class.

Mei Quanshun

PhD student in animal breeding

My research interests include genomic selection and machine learning in animal breeding.