getFeatureSpace: Get informative principal components

In IMB-Computational-Genomics-Lab/scPred: scPred - Single cell prediction using singular value decomposition and machine learning classification

Description

Given a prediction variable, finds a feature set of class-informative principal components. A Wilcoxon rank sum test is used to determine a difference between the score distributions of cell classes from the prediction variable.

Usage

getFeatureSpace(object, pVar, varLim = 0.01, correction = "fdr",
  sig = 0.05)

Arguments

object	An scPred or seurat object
pVar	Prediction variable corresponding to a column in metadata slot
varLim	Threshold to filter principal components based on variance explained.
correction	Multiple testing correction method. Default: false discovery rate. See p.adjust function
sig	Significance level to determine principal components explaining class identity

Value

An scPred object with two additional filled slots:

• features: A data frame with significant principal components the following information:

  • PC: Principal component

  • pValue: p-value obtained from Mann-Whitney test

  • pValueAdj: Adjusted p-value according to correction parameter

  • expVar: Explained variance by the principal component

  • cumExpVar: All principal components are ranked accoriding to their frequency of ocurrence and their variance explained. This column contains the cumulative variance explained across the ranked principal components

• pVar: Column name from metadata to use as the variable to predict using the informative principal components. Informative principal components are selected based on this variable.

Author(s)

Jos<c3><a9> Alquicira Hern<c3><a1>ndez

Examples

# Assign cell information to scPred object
# Cell information must be a data.frame with rownames as cell ids matching the eigendecomposed 
gene expression matrix rownames.

metadata(object) <- cellInfo

# Get feature space for column "cellType" in metadata slot

object <- getFeatureSpace(object = object, pVar = "cellType")

IMB-Computational-Genomics-Lab/scPred documentation built on Jan. 11, 2020, 7:37 a.m.