twoPhaseDE: Two-phase differential expression test for scRNA-seq data

In haowulab/SC2P: Two phase differential expression for single-cell RNA-seq

Description

This function performs differential expression analysis for singel-cell RNA-seq data in two phases: phase transition (probability of being expressed), and magnitude tuning (levels of expression once the gene is on). Currently the function only works for the comparison of two biological conditions.

Usage

twoPhaseDE(sc2p.obj, design, test.which, low.prob = 0.99, offset = c("SC2P", "sf"))

Arguments

sc2p.obj	An object of sc2pSet. Should be the result object from 'eset2Phase' function.
design	A character vector of variable names in pData(sc2p.obj) to provide regression covariates in DE test.
test.which	An integer for the index of the to-be-tested binary variable in argument 'design'.
low.prob	The cutoff used in classifying genes to Phase 2. Genes with posterior probability in Phase 2 greater than low.prob is considered in Phase 2.
offset	The normalization factor, or indicator for method to compute normalization factor. Valid entry includes: (1) a character, where 'SC2P' indicate to use the normalization procedure provided by SC2P. 'sf' uses the total counts. (2) a numeric vector, where each represent the normalization factor for a cell. The length of the vector must equal to the number of cells. (3) a numeric matrix of the same dimension as the input expression data, where each gene in each cell has a seperate normalization factor.

Details

est2Phase needs to be called before this function to estimate the probabilities of being in two states.

Value

A data frame with following columns:

p1	The Phase 2 proportion in Group 1.
p2	The Phase 2 proportion in Group 2.
Ph1.coef	Log Odds Ratio in testing DE in the form of phase change.
Ph1.pval	The p-value in testing DE in the form of phase change.
m1	Estiamted conditional mean expression in Group 1
m2	Estiamted conditional mean expression in Group 2
Ph2.coef	Conditional log2 fold change
Ph2.ci.lo	The lower end of 95% confidence interval for the conditional log fold change.
Ph2.ci.hi	The upper end of 95% confidence interval for the conditional log fold change.
Ph2.pval	The p-value in testing DE in the form of magnitude change within Phase 2.
logFC	Marginal log2 fold change

The data frame is sorted by gene names alphabetically.

Author(s)

Zhijin (Jean) Wu <[e-mail]>

See Also

est2Phase, topGene, visGene

Examples

## Not run: 
data(brain_scRNAseq)

## create ExpressionSet
colnames(Y) <- rownames(design)
phenoData <- new("AnnotatedDataFrame", data=design)
eset <- ExpressionSet(assayData=Y, phenoData=phenoData)

## estimate phases
data <- eset2Phase(eset)
data

##  DE test
de.sc2p <- twoPhaseDE(data, design="celltype", test.which=1, offset="sf")

## list top ranked DE genes
topGene(de.sc2p, phase=1)
topGene(de.sc2p, phase=2)
topGene(de.sc2p, phase="both")

## visualize top DE gene in phase 2
visGene(topGene(de.sc2p, 2)$Gene.name[1], data, group.name="celltype")

## End(Not run)

haowulab/SC2P documentation built on Feb. 28, 2021, 8:39 a.m.