normalize_data: Normalize expression data from a SummarizedExperiment or...

View source: R/stats.R

normalize_dataR Documentation

Normalize expression data from a SummarizedExperiment or matrix to be used in hipathia

Description

Transforms the rank of the SummarizedExperiment or matrix of gene expression to [0,1] in order to be processed by hipathia. The transformation may be performed in two different ways. If percentil = FALSE, the transformation is a re-scaling of the rank of the matrix. If percentil = TRUE, the transformation is performed assigning to each cell its percentil in the corresponding distribution. This option is recommended for distributions with very long tails.

Usage

normalize_data(
  data,
  sel_assay = 1,
  by_quantiles = FALSE,
  by_gene = FALSE,
  percentil = FALSE,
  truncation_percentil = NULL
)

Arguments

data

Either a SummarizedExperiment or a matrix of gene expression.

sel_assay

Character or integer, indicating the assay to be normalized in the SummarizedExperiment. Default is 1.

by_quantiles

Boolean, whether to normalize the data by quantiles. Default is FALSE.

by_gene

Boolean, whether to transform the rank of each row of the matrix to [0,1]. Default is FALSE.

percentil

Boolean, whether to take as value the percentil of each sample in the corresponding distribution.

truncation_percentil

Real number p in [0,1]. When provided, values beyond percentil p are truncated to the value of percentil p, and values beyond 1-p are truncated to percentil 1-p. By default no truncation is performed.

Details

This transformation may be applied either to the whole matrix (by setting by_gene = FALSE), which we strongly recommend, or to each of the rows (by setting by_gene = TRUE), allowing each gene to have its own scale.

A previous quantiles normalization may be applied by setting by_quantiles = TRUE. This is recommended for noisy data.

For distributions with extreme outlayer values, a percentil p may be given to the parameter truncation_percentil. When provided, values beyond percentil p are truncated to the value of percentil p, and values beyond 1-p are truncated to percentil 1-p. This step is performed before any other tranformation. By default no truncation is performed.

Value

Matrix of gene expression whose values are in [0,1].

Examples

data("brca_data")
trans_data <- translate_data(brca_data, "hsa")
exp_data <- normalize_data(trans_data)
exp_data <- normalize_data(trans_data, by_quantiles = TRUE,
truncation_percentil=0.95)


martahidalgo/hipathia documentation built on Jan. 12, 2023, 1:44 p.m.