README.md

In wiscstatman/EBSeq: An R package for gene and isoform differential expression analysis of RNA-seq data

About EBSeq

EBSeq is a Bioconductor package for processing sequence-based transcriptional profiling (RNA-Seq) data. It is based upon empirical Bayesian methodology, taking a matrix of expression data as input and returning transcript-specific vectors of local posterior probabilities that score various patterns of differential expression among the group labels associated with the samples.

The code here is a major version update of the original EBSeq method; it deploys some algorithmic changes to improve computational efficiency, especially in the multi-group setting.

This version 2.0 update of EBSeq is the topic of bioRxiv 10.1101/2020.06.19.162180 https://biorxiv.org/cgi/content/short/2020.06.19.162180v1

EBSeq.v2 package

Installation

You can install EBSeq.v2 from github with:

# install.packages("devtools")
devtools::install_github("wiscstatman/EBSeq")
library(EBSeq)

A synthetic two-group example

data(GeneMat)
str(GeneMat)

EBTest is the function for two-condition comparison. Most arguments of EBTest have default settings, expression data, condition label and size factors for normalization are required. If input data are normalized, then size factor for each sample is 1 and sizeFactors = rep(1,ncol(Data)), otherwise, we use median normalization to account for library size. GetPPMat yields the posterior probabilities being equivalent expressed and differential expressed.

Sizes = MedianNorm(GeneMat)
EBOut = EBTest(Data=GeneMat, Conditions=as.factor(rep(c("C1","C2"),each=5)),
       sizeFactors = Sizes)
PP = GetPPMat(EBOut)

The posterior probabilities in PP are specific to each gene (i.e., local) and to each pattern of equality/inequality of underlying means. The conditional false discovery rate of a list of reported genes is the mean of local probababilities of equivalent expression.

More than two conditions

Synthetic example:

data(MultiGeneMat)

EBMultiTest is the function considering differential expression patterns over more than two conditions. MultiOut$AllParti outputs the selected patterns, MultiPP gives the posterior probabilities for each expression patterns.

Conditions = c("C1","C1","C2","C2","C3","C3")
MultiSize = MedianNorm(MultiGeneMat)
MultiOut = EBMultiTest(MultiGeneMat,Conditions=Conditions,
                     sizeFactors=MultiSize)
patterns = MultiOut$AllParti
MultiPP = GetMultiPP(MultiOut)

More details and examples can be found in vignettes

Reference for the package can be found at "link to bioRxiv"

Analysis is available at https://github.com/wiscstatman/bigEB

Compatible with EBSeq.v1

The original EBSeq.v1 is embedded in the package, using the fast option to alternate between v2 and v1. By default fast is set to TRUE and EBSeq.v2 is called. To use EBSeq.v1, simply switch fast to FALSE

for example, the two group senario:

Sizes = MedianNorm(GeneMat)
EBOut = EBTest(Data=GeneMat, Conditions=as.factor(rep(c("C1","C2"),each=5)),fast = F,maxround = 5,
       sizeFactors = Sizes)
PP = GetPPMat(EBOut)

or more than two groups scenario:

Conditions = c("C1","C1","C2","C2","C3","C3")
MultiSize = MedianNorm(MultiGeneMat)
MultiOut = EBMultiTest(MultiGeneMat,Conditions=Conditions,fast = F,maxround = 5,
                     sizeFactors=MultiSize)
MultiPP = GetMultiPP(MultiOut)

Also support user-specified paritions

Like EBSeq.v1, user can use the argument AllParti to input self-specified partitions, that is, using prior knowledge, only considering a subset of total partitions.

Conditions = c("C1","C1","C2","C2","C3","C3")
PosParti=GetPatterns(Conditions)
PosParti
Parti=PosParti[-3,]
Parti

If we only consider patterns 1,2,4 and 5.

MultiSize = MedianNorm(MultiGeneMat)
MultiOut = EBMultiTest(MultiGeneMat,Conditions=Conditions,AllParti = Parti,
                     sizeFactors=MultiSize)
MultiPP = GetMultiPP(MultiOut)

wiscstatman/EBSeq documentation built on June 3, 2023, 7:34 a.m.