newSCESet: Create a new SCESet object.

In dynverse/scaterlegacy: Single-cell analysis toolkit for gene expression data in R

Description

Create a new SCESet object (the basic data container class in scater) from a supplied matrix of expression values, plus cell and feature metadata. The expression matrix have rows representing features (usually genes) and columns representing cells.

Usage

newSCESet(exprsData = NULL, countData = NULL, tpmData = NULL,
  fpkmData = NULL, cpmData = NULL, phenoData = NULL, featureData = NULL,
  experimentData = NULL, is_exprsData = NULL,
  cellPairwiseDistances = dist(vector()),
  featurePairwiseDistances = dist(vector()), lowerDetectionLimit = NULL,
  logExprsOffset = NULL)

Arguments

exprsData	expression data matrix for an experiment (features x cells)
countData	data matrix containing raw count expression values
tpmData	matrix of class "numeric" containing transcripts-per-million (TPM) expression values
fpkmData	matrix of class "numeric" containing fragments per kilobase of exon per million reads mapped (FPKM) expression values
cpmData	matrix of class "numeric" containing counts per million (CPM) expression values (optional)
phenoData	data frame containing attributes of individual cells
featureData	data frame containing attributes of features (e.g. genes)
experimentData	MIAME class object containing metadata data and details about the experiment and dataset.
is_exprsData	matrix of class "logical", indicating whether or not each observation is above the lowerDetectionLimit.
cellPairwiseDistances	object of class "dist" (or a class that extends "dist") containing cell-cell distance or dissimilarity values.
featurePairwiseDistances	object of class "dist" (or a class that extends "dist") containing feature-feature distance or dissimilarity values.
lowerDetectionLimit	the minimum expression level that constitutes true expression (defaults to zero and uses count data to determine if an observation is expressed or not).
logExprsOffset	numeric scalar, providing the offset used when doing log2-transformations of expression data to avoid trying to take logs of zero. Default offset value is 1.

Details

Scater requires that all data be housed in SCESet objects. SCESet extends Bioconductor's ExpressionSet class, and the same basic interface is supported. newSCESet() expects a single matrix of expression values of a nominated type to be provided, for example a matrix of counts or a matrix of transcripts-per-million values. There is a hierarchy applied to the expression data: counts > transcripts-per-million (tpm) > counts-per-million (cpm) > fragments-per-kilobase-per-million-mapped (fpkm) > generic expression values on the log2 scale (exprs). Data types higher in the higher are preferred. Data types lower in the hierarchy will be computed from values higher in the hierarchy - e.g. counts-per-million and expression values (as log2(cpm + offset)) will be computed from counts. Data types higher in the hierarchy will never be computed from types lower in the hierarchy (e.g. counts will never be computed from exprs values). At a minimum, an SCESet object will contain exprs values; these will be computed as log2(*pm + offset) values if a data type higher in the hierarchy is supplied as the expression matrix.

Per-feature and per-cell metadata can be supplied with the featureData and phenoData arguments, respectively. Use of these optional arguments is strongly encouraged.

Many methods are provided in the package that operate on SCESet objects.

Aside from the hierarchy of data types described above, scater is relatively agnostic with respect to data the nature of the expression values. Most frequently used values are feature counts or transcripts-per-million (tpm), but any valid output from a program that calculates expression values from RNA-Seq reads is supported. For example, expression values could also be values from a single cell qPCR run or some other type of assay.

In some cases it may be desirable to have both tpm and counts in an SCESet object. In such cases, expression matrices can be added to an SCESet object after it has been produced by using the set_exprs function to add the expression matrix to the SCESet object.

In many downstream functions it is most convenient if the 'exprs' values are on the log2-scale, so this is done by default.

Value

a new SCESet object

Examples

data("sc_example_counts")
data("sc_example_cell_info")
pd <- new("AnnotatedDataFrame", data = sc_example_cell_info)
example_sceset <- newSCESet(countData = sc_example_counts, phenoData = pd)
example_sceset

dynverse/scaterlegacy documentation built on Feb. 17, 2020, 5:07 a.m.