calculate_cvs: Compute mean expression level, standard deviation and...

In scFeatureFilter: A correlation-based method for quality filtering of single-cell RNAseq data

Description

Compute mean expression level, standard deviation and coefficient of variation (CV) of each feature (i.e. gene or transcript) in the supplied data. Filter features with high proportion of 0 expression.

Usage

calculate_cvs(data, max_zeros = 0.75, sce_assay = NULL)

Arguments

data	A data frame, a matrix or a SingleCellExperiment object. If data frame or matrix, it should contain expression values for each gene as rows, and expression values for the cells as columns.
max_zeros	A number between 0 and 1 indicating the maximum proportion of zero expression values allowed per row. Features with a higher proportion of 0 will be discarded.
sce_assay,	if data is an SingleCellExperiment object, sce_assay should be one of names(assays(<SingleCellExperiment>)).

Details

Before CV computation, the function removes all rows that have a proportion of zeros above the specified threshold. Genes with many 0s are poorly informative, and would bias the later correlations. Removing them also prevents division by zero when calculating CVs.

The data provided must cell/sample names as column names. Feature name can be given either in the first column or as row names.

In the output, mean, standard deviation and CV are incorporated as new columns in the data frame, named mean, sd and cv.

Value

A data frame, containing the filtered data with additional columns: mean, standard deviation and cv values for each row.

Examples

expMat <- matrix(
    c(1, 1, 1,
      1, 2, 3,
      0, 1, 2,
      0, 0, 2),
    ncol = 3, byrow = TRUE, dimnames = list(paste("gene", 1:4), paste("cell", 1:3))
)
calculate_cvs(expMat)
calculate_cvs(expMat, max_zeros = 0.5)

scFeatureFilter documentation built on Nov. 8, 2020, 7:49 p.m.