Examples

The following chunks of codes demonstrate how functions are used to perform assignment for different data types and tasks. Please change file and variable names accordingly, and use different argument values for your needs.

#Import a GENEPOP file
geninfile <- read.Genepop("YourGenepopFile.txt", pop.names=c("pop_A","pop_B","pop_C"))

#Reduce low variance loci (optional)
geninfile_rd <- reduce.allele(geninfile, p=0.95)

#Perform Monte-Carlo cross-validation, with using subsets of high Fst loci as training loci, LDA to build predictive models
assign.MC(geninfile_rd, train.inds=c(0.5, 0.7, 0.9), train.loci=c(0.1, 0.25, 0.5, 1), loci.sample="fst",
          iterations=30, model="lda", dir="ResultFolder_MC/")

#Calculate assignment accuracy
accuRes_MC <- accuracy.MC(dir="ResultFolder_MC/")

#Make assignment accuracy boxplot
accuracy.plot(accuRes_MC, pop=c("all","pop_A","pop_B","pop_C"))

#Perform K-fold cross-validation 
assign.kfold(geninfile_rd, k.fold=c(3, 4, 5), train.loci=c(0.1, 0.25, 0.5, 1), loci.sample="fst",
             model="lda", dir="ResultFolder_KF/")

#Make membership probability plot
membership.plot(dir="ResultFolder_KF/")

#Identify informative loci
check.loci(dir="ResultFolder_MC/", top.loci=20)

#Import a STRUCTURE file
strinfile <- read.Structure("YourStructureFile.txt")

#Concatenate genepop and non-genetic data
conData <- compile.data(strinfile, "OtherFeature.csv")

#Perform Monte-Carlo cross-validation, with subsets of random loci plus additional features as training data. The whole data is scaled (scaled=TRUE) before performing PCA and cross-validation. Each PC summarizes variance of genetic and non-genetic data (pca.method="mixed").
assign.MC(conData, train.inds=c(30, 40, 50), train.loci=c(0.1, 0.25, 0.5, 1), loci.sample="random",
          iterations=30, pca.method="mixed", scaled=TRUE, model="svm", dir="ResultFolder_MC/")

#Calculate assignment accuracy
accuRes_MC <- accuracy.MC(dir="ResultFolder_MC/")

#Make assignment accuracy boxplot
accuracy.plot(accuRes_MC, pop=c("all","pop_A","pop_B","pop_C"))

#See Example 1 for performing K-fold cross-validation and making membership probability plot.

#Import non-genetic data set. Note that this data set should include a population label in the last column
infile <- read.csv("YourNonGenetics.csv", header=TRUE)

#Convert your sample ID and population name to factor data type, if they are numeric (optional)
infile$ID <- as.factor(infile$ID)
infile$pop <- as.factor(infile$pop)

#Perform Monte-Carlo cross-validation, with transformed data (pca.method=TRUE, transform original data to PCs for dimentionality reduction) as new features to build decision tree predictive models 
assign.MC(infile, train.inds=c(0.5, 0.7, 0.9), iterations=30, pca.method=TRUE, model="tree",
          dir="ResultFolder_MC/")

#Calculate assignment accuracy
accuRes_MC <- accuracy.MC(dir="ResultFolder_MC/")

#Make assignment accuracy boxplot
accuracy.plot(accuRes_MC, pop=c("all","pop_A","pop_B","pop_C"))

#See Example 1 for performing K-fold cross-validation and making membership probability plot.

#Import genetic data of known individuals
genin_known <- read.Genepop("YourGenepop_Known.txt", pop.names=c("pop_A","pop_B","pop_C"))

#Concatenate genetic and non-genetic data of known individuals
conData_known <- compile.data(genin_known, "OtherFeatures_Known.csv")

#Import genetic data of unknown individuals. Note that this Genepop file should only have one "pop" label
genin_unknown <- read.genpop("YourGenepop_Unknown.txt")

#Concatenate genetic and non-genetic data of unknown individuals
conData_unknown <- compile.data(genin_unknown, "OtherFeatures_Unknown.csv")

#Perform assignment test using original non-genetic data as features (pca.method="original") to build random forest (tree number=50) predictive model
assign.X(conData_known, conData_unknown, pca.method="original", model="randomForest", ntree=50,
         dir="ResultFolder/")