feature_select {imsig}

Feature selection of signature genes

Description

ImSig genes were designed to be co-expressed in tissue transcriptomic data. However, depending on the dataset some of the genes may not co-express with the dominant module. In order to remove such deviant genes, a feature selection can be carried out based on correlation. This function removes genes that exhibit a poor correlation (less than the defined r value) with the dominant ImSig module. This step of feature selection is recommended to enrich the prediction of relative abundance of immune cells.

Usage

feature_select(exp, r = 0.6)

Arguments

exp

Dataframe of transcriptomic data (natural scale) containing genes as rows and samples as columns. Note: Gene names should be set as row names and duplicates are not allowed. Missing values are not allowed within the expression matrix. Check example- head(example_data): example_data.

r

Use a value between 0 and 1. Default is 0.6. This is a user defined correlation cut-off to perform feature selection. To get an idea of what cut-off to use check the results of (gene_stat) and choose a cut-off that displays high median correlation and maintains a high proportion of genes after feature selection.

Value

Returns a list of 'feature selected' genes based on the set r value.

Examples

feature_select (exp = example_data, r = 0.7)