extract_top_genes: Select top highly expressed genes.

In angelesarzalluz/scfilters: A correlation-based method for quality filtering of single-cell RNAseq data

Description

Selects the group of genes in the dataset that will be considered more highly expressed, the top window, using the filtering method chosen by the user.

Usage

extract_top_genes(dataset, method = c("window_size", "min_expression",
  "mean_expression"), parameter)

Arguments

dataset	A data frame, containing genes as rows and cells as columns, and where the mean expression value for each gene has been added as a column.
method	A string indicating the method to use when creating the top window. If no method is indicated, "window_size" will be used.
parameter	An integer. Indicates the numeric parameter to use in the previously chosen method.

Details

There are three selection methods available:

• "window_size": genes are ranked by mean expression, and a subset of the size indicated in parameter is selected from the top.

• "min_expression": genes where all expression values are above a minimum expression threshold indicated in parameter are selected.

• "mean_expression": the mean column is checked, and all genes with mean expression above the threshold indicated in parameter are selected.

There are no restrictions to the parameter argument, however, the value should be coherent with the characteristics of the data set provided and the chosen method.

In general, it is adviseable to avoid generating top windows much larger than 250 genes, to prevent excessively long computation time as well as to preserve the quality of the analysis, as the top window should only include a subset of reliable values. As a rule, the bigger the top window is, the more likely is that the reliability of the values is compromised, given the characteristics of single cell RNA sequencing data.

Value

A list with two elements, both data frames: the generated top window, and the rest of the genes.

angelesarzalluz/scfilters documentation built on May 10, 2019, 11:46 a.m.