ABSSeqlm: Differential expression analysis for complex desgin.

In ABSSeq: ABSSeq: a new RNA-Seq analysis method based on modelling absolute expression differences

Description

This function performs a default analysis by calling, in order, the functions: normalFactors, aFoldcomplexDesign,

Usage

ABSSeqlm(object, design, condA, condB = NULL, lmodel = TRUE,
  preval = 0.05, qforkappa = 0, adjmethod = "BH", scale = FALSE,
  quiet = FALSE, ...)

Arguments

object	a ABSDataSet object (not need 'groups' information).
design	a numeric matrix for expriment, with samples and factors in rows and colnums, respectively. Design respresents the satuarated model.
condA	a vector of factors for DE analysis, which could be redundant, see aFoldcomplexDesign.
condB	a vector of factors for DE analysis, which could be redundant, default is null, if not provide, the DE analysis will switch to assess difference across factors in condA (analysis of variance). If provide, DE analysis will focus on contrast between condB and condA (condB-condA). See aFoldcomplexDesign. The unique factors in condA+condB represents the reduced model.
lmodel	switch of fit linear model from limma-lmFit under design, default is TRUE. If TRUE, a gene-specific residual varaince will be estimated from (satuarated model - reduced model). Satuarated model includes all factors in design matrix and reduced model includes factors in condA+condB. if satuarated model == reduced model, the DE analysis performs pairwise comparison or one-way analysis of variance. See aFoldcomplexDesign.
preval	parameter for aFoldcomplexDesign, prior value for controlling of variance scale in case over-scaled, default is 0.05,
qforkappa	parameter for aFoldcomplexDesign, quantile for estimating kappa(>=qforkappa), default is 0 (no trimming of data).
adjmethod	defualt is 'BH', method for p-value adjusted, see p.adjust.methods for details
scale	switch for scaling fold change according to common SD under log2 transformation, default is FALSE.
quiet	default is FALSE, whether to print messages at each step
...	parameters passed to lmFit in limma

Details

This function uses a linear model (limma-lmFit) to infer DE under complex design.

Value

a result table with additional elements, including: basemean, log of basemean, foldChange, shrinked (expression level and gene-specific) log2 of fold-change, B - A, or (SDs under log2 for analysis of variance) pvalue, pvalue from NB distribution model, p.adj, adjuested p-value used p.adjust method. scaledlogFC, scaled logFC if scale=TRUE.

Author(s)

Wentao Yang

References

Wentao Yang, Philip Rosenstiel & Hinrich Schulenburg: ABSSeq: a new RNA-Seq analysis method based on modelling absolute expression differences

Examples

data(simuN5)
groups=factor(simuN5$groups)
obj <- ABSDataSet(counts=simuN5$counts)
design <- model.matrix(~0+groups)
res <- ABSSeqlm(obj,design,condA=c("groups0"),condB=c("groups1"))
head(res)

ABSSeq documentation built on Nov. 8, 2020, 5:07 p.m.