DEAnalysisMAST: DEAnalysisMAST

In DWLS: Gene Expression Deconvolution Using Dampened Weighted Least Squares

DEAnalysisMAST

Description

Perform DE analysis using MAST. Dampened weighted least squares (DLWS) is an estimation method for gene expression deconvolution, in which the cell-type composition of a bulk RNA-seq data set is computationally inferred. This method corrects common biases towards cell types that are characterized by highly expressed genes and/or are highly prevalent, to provide accurate detection across diverse cell types. To begin, the user must input a bulk RNA-seq data set, along with a labeled representative single-cell RNA-seq data set that will serve to generate cell-type-specific gene expression profiles. Ideally, the single-cell data set will contain cells from all cell types that may be found in the bulk data. DWLS will return the cell-type composition of the bulk data. First, solve OLS then use the solution to find a starting point for the weights. Next, the dampened weighted least squares is performed. The weights are iterated until convergence then the dampening constant for weights is found using cross-validation (with decreasing step size for convergence).

DWLS captures ISC composition changes across conditions. One of the most important applications of deconvolution methods is in the identification of cell-type composition variations across conditions.

Note: The function uses solveDampenedWLSj() and findDampeningConstant().

Usage

DEAnalysisMAST(scdata, id, path)

Arguments

scdata	The gene expression datafarme
id	The unique identities within the data
path	The path for the RData results

Value

matrix The resulting matrix is a gene by cell-type signature matrix. The cell-type signature matrix is constructed using a representative single-cell data set, such that all cell types expected in the bulk data are also represented in the single-cell data (the converse need not be true). The single-cell data is first clustered to reveal its constituent cell types.The function will return 3 different files: an RData file, an rds file, and a csv file.

Examples

#dataSC
#url <- "https://github.com/sistia01/DWLS/raw/main/inst/extdata/dataSC.RData"
#dest <- "data/dataSC.RData"
#load(download.file(url, tempfile(data/dataSC.RData))
#load("dataSC.RData")
#SOLUTION
load(system.file("extdata", "dataSC.RData", package = "DWLS"))

#dataBulk
url <- "https://github.com/sistia01/DWLS/raw/main/inst/extdata/dataBulk.RData"
dest <- "data/dataBulk.RData"
load(download.file(url, tempfile(dest)))
#load("data/dataBulk.RData")
load(system.file("extdata", "dataBulk.RData", package = "DWLS"))

#labels
#url <- "https://github.com/sistia01/DWLS/raw/main/inst/extdata/labels.RData"
#dest <- "data/labels.RData"
#download.file(url, dest)
#load("data/labels.RData")
load(system.file("extdata", "labels.RData", package = "DWLS"))

#data('trueLabels', package = "DWLS")
#url <- "https://github.com/sistia01/DWLS/raw/main/inst/extdata/trueLabels.RData"
#dest <- "data/trueLabels.RData"
#download.file(url, dest)
#load("data/trueLabels.RData")
load(system.file("extdata", "trueLabels.RData", package = "DWLS"))

labels<-trueLabels

#Old Method
#load("data/dataBulk.RData") #read in bulk data for WT1 (control condition #1)
#load("data/labels.RData") #read in single-cell labels from clustering
#data('dataSC_3', package = "DWLS")
#dataSC <- dataSC_3

labels<-trueLabels

#Old Method
#load("data/dataBulk.RData") #read in bulk data for WT1 (control condition #1)
#load("data/labels.RData") #read in single-cell labels from clustering
#data('dataSC_3', package = "DWLS")
#dataSC <- dataSC_3

labels<-trueLabels

#Change to real labels
newcat<-c("NonCycISC","CycISC","TA","Ent","PreEnt","Goblet","Paneth","Tuft",
 "EE")
for (i in 1:length(newcat)){
  labels[which(labels==(i-1))]<-newcat[i]
  }

#Run deconvolution
#Results are in inst/extdata/results folder -- run on local
#Example code below
Mast_test <- DEAnalysisMAST(dataSC, labels, "inst/extdata/results")

DWLS documentation built on May 24, 2022, 5:07 p.m.