processNanostringData: Process NanoString nCounter gene expression data.

In calebclass/NanoTube: An Easy Pipeline for NanoString nCounter Data Analysis

Process NanoString nCounter gene expression data.

Description

This function reads in a zip file or folder containing multiple .rcc files (or a txt/csv file containing raw count data), and then optionally conducts positive control normalization, background correction, and housekeeping normalization.

Usage

processNanostringData(
  nsFiles,
  sampleTab = NULL,
  idCol = NULL,
  groupCol = NULL,
  replicateCol = NULL,
  normalization = c("nSolver", "RUVIII", "RUVg", "none"),
  bgType = c("threshold", "t.test", "none"),
  bgThreshold = 2,
  bgProportion = 0.5,
  bgPVal = 0.001,
  bgSubtract = FALSE,
  n.unwanted = NULL,
  RUVg.drop = 0,
  housekeeping = NULL,
  skip.housekeeping = FALSE,
  includeQC = FALSE,
  sampIds = NULL,
  output.format = c("ExpressionSet", "list"),
  logfile = ""
)

Arguments

nsFiles	file path (or zip file) containing the .rcc files, or multiple directories in a character vector, or a single text/csv file containing the combined counts, or .rcc files in a character vector.
sampleTab	.txt (tab-delimited) or .csv (comma-delimited) file containing sample data table (optional, default NULL)
idCol	the column name of the sample identifiers in the sample table, which should correspond to the column names in the count table (default NULL: will assume the first column contains the sample identifiers)
groupCol	the column name of the group identifiers in the sample table.
replicateCol	the column name of the technical replicate identifiers (default NULL). Multiple replicates of the same sample will have the same value in this column. Replicates are used to improve normalization performance in the "RUVIII" method; otherwise they are averaged.
normalization	If "nSolver" (default), continues with background, positive control, and housekeeping control normalization steps to return a NanoStringSet of normalized data. If "RUVIII", runs RUV normalization using controls, housekeeping genes and technical replicates. If "RUVg", runs RUV normalization using housekeeping genes. If "none", returns a NanoStringSet with the raw counts, suitable for running NanoStringDiff.
bgType	(Only if normalization is not "none") Type of background correction to use: "threshold" sets a threshold for N standard deviations above the mean of negative controls. "t.test" conducts a one-sided t test for each gene against all negative controls. "none" to skip background removal
bgThreshold	If bgType=="threshold", number of sd's above the mean to set as threshold for background correction.
bgProportion	If bgType=="threshold", proportion of samples that a gene must be above threshold to be included in analysis.
bgPVal	If bgType=="t.test", p-value threshold to use for gene to be included in analysis.
bgSubtract	Should calculated background levels be subtracted from reported expressions? If TRUE, will subtract mean+numSD*sd of the negative controls from the endogenous genes, and then set negative values to zero (default FALSE)
n.unwanted	The number of unwanted factors to use (for RUVIII or RUVg normalization only). If NULL (default), the maximum possible value will be identified and used.
RUVg.drop	The number of singular values to drop for RUVg normalization (see RUVSeq::RUVg)
housekeeping	vector of genes (symbols or accession) to use for housekeeping correction ("nCounter" or "RUVg" normalization). If NULL, will use genes listed as "Housekeeping" under CodeClass.
skip.housekeeping	Skip housekeeping normalization? (default FALSE)
includeQC	Should we include the QC from the .rcc files? This can cause errors, particularly when reading in files from multiple experiments.
sampIds	a vector of sample identifiers, important if there are technical replicates. Currently, this function averages technical replicates. sampIds will be extracted from the replicateCol in the sampleTab, if provided.
output.format	If "list", will return the normalized (optional) and raw expression data, as well as various QC and relevant information tables. If "ExpressionSet" (default), will convert to an n*p ExpressionSet, with n rows representing genes and p columns representing samples. ExpressionSet objects are required for some steps, such as runLimmaAnalysis.
logfile	a filename for the logfile (optional). If blank, will print warnings to screen.

Value

An list or ExpressionSet containing the raw and/or normalized counts, dictionary, and sample info if provided

Examples

example_data <- system.file("extdata", "GSE117751_RAW", package = "NanoTube")
sample_data <- system.file("extdata", "GSE117751_sample_data.csv", 
                           package = "NanoTube")

# Process NanoString data from RCC files present in example_data folder.
# Use standard nCounter normalization, removing genes that do
# pass a t test against negative control genes with p < 0.05. Return the
# result as an "ExpressionSet".

dat <- processNanostringData(nsFiles = example_data,
                             sampleTab = sample_data, 
                             groupCol = "Sample_Diagnosis",
                             normalization = "nSolver",
                             bgType = "t.test", bgPVal = 0.01,
                             output.format = "ExpressionSet")

# Load NanoString data from a csv file (from NanoString's RCC Collector tool,
# for example). Skip normalization by setting 'normalization = "none"'.

csv_data <- system.file("extdata", "GSE117751_expression_matrix.csv", 
                        package = "NanoTube")
dat <- processNanostringData(nsFile = csv_data,
                              sampleTab = sample_data, 
                              idCol = "GEO_Accession", 
                              groupCol = "Sample_Diagnosis",
                              normalization = "none")
                              
# Load NanoString data from RCC files, using a threshold background level for
# removing low-expressed genes. Also, specify which genes to use for 
# housekeeping normalization. Save the result in "list" format (useful for
# some QC functions) instead of an "ExpressionSet".

dat <- processNanostringData(nsFiles = example_data,
                             sampleTab = sample_data, 
                             groupCol = "Sample_Diagnosis",
                             normalization = "nSolver",
                             bgType = "threshold", 
                             bgThreshold = 2, bgProportion = 0.5,
                             housekeeping = c("TUBB", "TBP", "POLR2A", 
                                              "GUSB", "SDHA"),
                             output.format = "list")

calebclass/NanoTube documentation built on Nov. 21, 2023, 12:31 p.m.