DaMiR.FReduct: Remove highly correlated features, based on...

In BioinfoMonzino/DaMiRseq: Data Mining for RNA-seq data: normalization, feature selection and classification

Description

This function allows the user to remove highly correlated features.

Usage

DaMiR.FReduct(data, th.corr = 0.85, type = c("spearman", "pearson"))

Arguments

data	A transposed data frame or matrix of normalized expression data. Rows and Cols should be, respectively, observations and features
th.corr	Feature-per-feature correlation threshold; default is 0.85
type	Type of correlation metric to be applied; default is "spearman"

Details

This function produces an absolute correlation matrix that it is then used to reduce pair-wise correlations. When two features present a correlation higher than that defined by the user in th.corr argument, the function, first, calculates the mean absolute correlation of each feature and, then, removes the feature with the largest mean absolute correlation.

Value

An expression matrix without highly correlated features.

Author(s)

Mattia Chiesa, Luca Piacentini

See Also

rcorr, findCorrelation

Examples

# use example data:
data(data_reduced)
# reduce the number of features:
data_Reduced <- DaMiR.FReduct(data_reduced,
th.corr = 0.75, type = "pearson")

BioinfoMonzino/DaMiRseq documentation built on Aug. 22, 2021, 3:11 p.m.