ICE_cv: Cross validation function for ICE imputation accuracy

In bioinformatist/ICE: ICE: Impute CircRNA Expression

Description

Function to obtain accuracy parameters: correlation coefficient, P-value and RMSE of imputation model

Usage

ICE_cv(
  train.pcg,
  train.circ,
  gene.index,
  num = 50,
  method = "KNN",
  folds = 10,
  filter = TRUE,
  cor.method = "spearman",
  ...
)

Arguments

train.pcg	training protein coding dataset, which should be a numeric matrix with with row names indicating samples, and column names indicating protein coding gene IDs.
train.circ	training circRNA expression dataset, which should be a numeric matrix with row names indicating samples, and column names indicating circRNA IDs.
gene.index	either gene name (character) or index (column number) of circRNA to be imputed.
num	number of informative protein coding genes to be used in constructing imputation model. Default is 100 genes.
method	method for imputation, either "RF" for random forests, "KNN" for K-nearest neighbor or "SVM" for support vector machines.
folds	number specifying folds (k) of cross validation to obtain imputation accuracy. Default is k=10.
filter	if training datasets should be filtered? Default is TRUE, but should be set to FALSE when called in loop.
cor.method	a character string indicating which correlation coefficient is to be used for the correlation test. Default is "spearman", could also be "pearson".
...	optional parameters that can be passed on to the machine-learning functions RF (randomForest), KNN (knn.reg) or SVM(svm)

Value

a matrix with three values corresponding to correlation coefficient, P-value of the fit and root mean squared error (RMSE).

Examples

data(iMIRAGE.datasets)
ICE_cv(GA.pcg, GA.mir, gene.index="hsa-let-7c", method="KNN", num=100)
ICE_cv(GA.pcg, GA.mir, gene.index=25, method="KNN", num=100)

bioinformatist/ICE documentation built on July 5, 2020, 12:20 a.m.