Package 'XDE' reference manual

Title:	XDE: a Bayesian hierarchical model for cross-study analysis of differential gene expression
Description:	Multi-level model for cross-study detection of differential gene expression.
Authors:	R.B. Scharpf, G. Parmigiani, A.B. Nobel, and H. Tjelmeland
Maintainer:	Robert Scharpf <[email protected]>
License:	LGPL-2
Version:	2.53.0
Built:	2024-11-27 05:11:33 UTC
Source:	https://github.com/bioc/XDE

Indicator for running a MCMC burnin

Description

When TRUE, log files from MCMC chains are not written to file. When FALSE, log files are written for every parameter by default.

Usage

burnin(object)
burnin(object)

Arguments

object

An object of class XdeParameter

Value

logical

Author(s)

R. Scharpf

Examples

## Not run: 
data(expressionSetList)
params <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
              esetList=expressionSetList)

##the replacement method for burnin is called for its side effect of
##providing default values of storing MCMC chains
output(params)[2:22]
burnin(params) <- FALSE
output(params)[2:22]
burnin(params) <- TRUE
output(params)[2:22]

## End(Not run)
## Not run: 
data(expressionSetList)
params <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
              esetList=expressionSetList)

##the replacement method for burnin is called for its side effect of
##providing default values of storing MCMC chains
output(params)[2:22]
burnin(params) <- FALSE
output(params)[2:22]
burnin(params) <- TRUE
output(params)[2:22]

## End(Not run)

Calculate the posterior average for indicators of concordant and discordant differential expression

Description

This function calculates the posterior average for indicators of concordant and discordant differential expression from the saved log files. See details.

Usage

calculatePosteriorAvg(object, NCONC=2, NDIFF=1, burnin=0)
calculatePosteriorAvg(object, NCONC=2, NDIFF=1, burnin=0)

Arguments

`object`	Object of class `XdeMcmc`
`NCONC`	Integer: number of studies for which the gene must be differentially expressed (in the same direction) to be classified as concordant differential expression
`NDIFF`	Integer: number of studies for which a gene must be up- or down-regulated to be classified as differentially expressed. It is the union of concordant and discordant differential expression.
`burnin`	Integer: number of MCMC iterations for the burnin. Posterior means are computed from the MCMC samples following burnin.

Details

For each iteration,

1. calculate the sign of delta * Delta

2. For each gene, compute the number of positive signs (P) and the number of negative signs (N) (a G x 2 matrix, where G is the number of genes in common across all studies). P + N is <= S, where S is the number of studies.

3. for a given gene, the discordant indicator is simply when P * N is nonzero.

4. The concordant indicator requires P * N = 0 AND P + N >= NCONC, where NCONC is specified by the user.

5. differential expression is simply | P | + | N | >= NDIFF. By default, NDIFF is 1 but can be user-specified.

The posterior average is then computed from the mean over all MCMC iterations.

Value

A G x 3 matrix.

Author(s)

Empirical starting values for the MCMC

Description

Empirical starting values for the MCMC are based on data in objects of class ExpressionSetList

Usage

empiricalStart(object, zeroNu = FALSE, phenotypeLabel, one.delta=FALSE, T_THRESH=4)
empiricalStart(object, zeroNu = FALSE, phenotypeLabel, one.delta=FALSE, T_THRESH=4)

Arguments

`object`	An object of class `ExpressionSetList`
`zeroNu`	Logical: if TRUE, the nu in the Bayesian model are not modeled – set to zero and not updated in the MCMC. Setting zeroNu to TRUE should be regarded as experimental
`phenotypeLabel`	character: binary phenotype. phenotypeLabel must be in the varLabels of each ExpressionSet object
`one.delta`	delta in the Bayesian model is a gene-specific indicator for differential expression. If one.delta is FALSE, we assume that a gene can be differentially expressed in a subset of studies. When TRUE, we assume that a gene is differentially expressed in all studies or in none.
`T_THRESH`	A threshold of t-statistics (calculated row-wise for each study) for determining starting values of the differential expression indicator, delta.

Value

A list containing starting values for the MCMC that are derived from empirical estimates of the data.

Author(s)

R. Scharpf

Examples

library(XDE)
data(expressionSetList)
eList <- studyCenter(expressionSetList)
empirical <- empiricalStart(eList, phenotypeLabel="adenoVsquamous", T_THRESH=3)
##By default, initial values for the MCMC are sampled from the prior
##when initializing an object of class XdeParamater
params <- new("XdeParameter", esetList=eList,
	      phenotypeLabel="adenoVsquamous", one.delta=FALSE, burnin=TRUE)
##The initial values can be replaced by empirical values as follows:
firstMcmc(params) <- empirical
library(XDE)
data(expressionSetList)
eList <- studyCenter(expressionSetList)
empirical <- empiricalStart(eList, phenotypeLabel="adenoVsquamous", T_THRESH=3)
##By default, initial values for the MCMC are sampled from the prior
##when initializing an object of class XdeParamater
params <- new("XdeParameter", esetList=eList,
	      phenotypeLabel="adenoVsquamous", one.delta=FALSE, burnin=TRUE)
##The initial values can be replaced by empirical values as follows:
firstMcmc(params) <- empirical

Example of ExpressionSetList

Description

Object of class ExpressionSetList containing three studies. Each element in the list is an ExpressionSet

Usage

data(expressionSetList)data(expressionSetList)

Details

Parmigiani et al. (2004) performed a cross-study analysis of three lung cancer studies. The studies used in this analysis were merged by UniGene identifiers to obtain a set of 3,171 gene. The R experiment data package lungExpression that was developed to facilitate the reproducibility of this analysis contains the three studies as ExpressionSets. Here, we take a random sample of 500 features from one study (the "stanford" study), and split this study into three artificial studies that each contain 4 squamous carcinomas and 3 adenocarcinomas. The three artificial studies are then used to create an instance of the ExpressionSetList class.

See Garber et al. (2001) for the raw data and description of the stanford study.

Source

The experiment data package lungExpression (www.bioconductor.org)

References

Parmigiani et al. (2004) A cross-study comparison of gene expression studies for the molecular classification of lung cancer, Clin Cancer Res, 10(9): 2922-2927

Garber et al. (2001) Diversity of gene expression in adenocarcinoma of the lung, PNAS, 98:13784-13789

Examples

data(expressionSetList)
data(expressionSetList)

A class for containing a list of ExpressionSets

Description

Each element in the list must be a valid ExpressionSet. The featureNames must be identical for each ExpressionSet.

Objects from the Class

Objects can be created by calls of the form new("ExpressionSetList", ...).

Slots

.Data:: Object of class "list"

Extends

Class "list", from data part. Class "vector", by class "list", distance 2. Class class.AssayData, by class "list", distance 2.

Methods

"[": signature(x = "ExpressionSetList") Subsets each ExpressionSet element in the list.
coerce: signature(from = "list", to = "ExpressionSetList") Coerces a list of ExpressionSet objects to an object of class ExpressionSetList. The validityMethod for the ExpressionSetList class will return an error if the featureNames for each ExpressionSet are not identical.
dim: signature(x="ExpressionSetList") applies dim to each element of the list.
featureNames: signature(object = "ExpressionSetList") Accessor for the featureNames
geneCenter: signature(object = "ExpressionSetList") See geneCenter
lapply: signature(object="ExpresssionSetList") Coerces instance of ExpressionSetList to a list and does lapply on the list. Returns an object of class ExpressionSetList
nSamples: signature(x = "ExpressionSetList") Numerical vector giving the number of samples in each ExpressionSet
nrow: signature(x = "ExpressionSetList") Numerical: number of features or genes
pData: signature(object = "ExpressionSetList") returns a list of data.frames. The elements of the list correspond to the studies in the ExpressionSetList object.
phenotype: signature(object="ExpressionSetList", varLabel="character") Accessor for the clinical variable. Assumes that the clinical variable has the same name in each study.
standardizeSamples: signature(object = "ExpressionSetList") See standardizeSamples
studyCenter: signature(object = "ExpressionSetList") See studyCenter
zeroNu: signature(object = "ExpressionSetList") See zeroNu.

Author(s)

R. Scharpf

Examples

	showClass("ExpressionSetList")
	data(expressionSetList)
showClass("ExpressionSetList")
	data(expressionSetList)

Methods for ExpressionSetList

Description

Methods for objects of class ExpressionSetList.

Usage

  phenotype(object, varLabel)
phenotype(object, varLabel)

Arguments

`object`	A `ExpressionSetList`.
`varLabel`	`character`. Name of the clinical variable.

Value

phenotype returns a matrix of the clinical variable where each column is a study. We require that the clinical variable have the same name in each study (each element of the ExpressionSetList object) and that the clinical variable is binary with values 1 or 0.

Values for the first MCMC iteration

Description

Accessor method for the values of the first MCMC iteration

Usage

firstMcmc(object)
firstMcmc(object)

Arguments

object

An object of class XdeParameter

Value

Returns a list of the values to be used in the first iteration of the MCMC.

Author(s)

R. Scharpf

Examples

data(expressionSetList)
params <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
              esetList=expressionSetList)
str(firstMcmc(params))
data(expressionSetList)
params <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
              esetList=expressionSetList)
str(firstMcmc(params))

Center the expression values for each gene in a study to zero

Description

Mean centers the genes for each study in a list

Usage

geneCenter(object)
geneCenter(object)

Arguments

object

Object of class ExpressionSetList

Value

Object of class ExpressionSetList

Author(s)

R. Scharpf

Examples

  data(expressionSetList)
  centered <- geneCenter(expressionSetList)
data(expressionSetList)
  centered <- geneCenter(expressionSetList)

Accessor for hyperparameters of the Bayesian model

Description

Accessor and replacement methods for hyperparameters of the Bayesian model are provided

Usage

hyperparameters(object)
hyperparameters(object)

Arguments

object

An object of class XdeParameter

Details

See the XdeParameterClass vignette for a more detailed discussion. The default values provided when initializing an object of class XdeParameter works well in most instances.

Value

A numerical vector

Author(s)

R. Scharpf

References

R. Scharpf et al., A Bayesian Model for Cross-Study Differential Gene Expression, Technical Report 158, Johns Hopkins University, Department of Biostatistics, 2007

Examples


data(expressionSetList)
xlist <- new("XdeParameter", esetList=expressionSetList, phenotypeLabel="adenoVsquamous")
hyperparameters(xlist)

data(expressionSetList)
xlist <- new("XdeParameter", esetList=expressionSetList, phenotypeLabel="adenoVsquamous")
hyperparameters(xlist)

Number of MCMC iterations

Description

Number of MCMC iterations

Usage

iterations(object)
iterations(object)

Arguments

object

An object of class XdeParameter or XdeMcmc.

Details

For an object of class XdeParameter, iterations specifies the total number of MCMC iterations. Note that by setting the thin parameter to a value greater than 1, the number of MCMC iterations will be greater than the number of saved MCMC iterations (saved iterations = iterations / thin).

For an object of class XdeMcmc (a class that stores output from the MCMC), iterations specifies the number of iterations that were saved.

The replacement method is only defined for the XdeParameter class. The class XdeMcmc is meant to reflect the information in an already run chain, whereas XdeParameter is a class for parameterizing the Bayesian model that has not yet been fit.

Value

An integer

Author(s)

R. Scharpf

MCMC values for the last iteration

Description

MCMC values for the last iteration. Useful if more iterations are needed.

Usage

  lastMcmc(object)
lastMcmc(object)

Arguments

object

Object of class XdeMcmc

Value

An environment.

Author(s)

R. Scharpf

Examples

  ## Not run: 
    data(expressionSetList)
    xparam <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
                  esetList=expressionSetList)
    iterations(xparam) <- 10  
    fit <- xde(xparam, esetList=expressionSetList)
    ##Do more iterations and use a different seed
    firstMcmc(xparam) <- lastMcmc(fit)
    seed(xparam) <- 97814
    fit2 <- xde(xparam, esetList=expressionSetList)

    ##Or
    fit2 <- xde(xparam, esetList=expressionSetList, outputMcmc=fit)
  
## End(Not run)
## Not run: 
    data(expressionSetList)
    xparam <- new("XdeParameter", phenotypeLabel="adenoVsquamous",
                  esetList=expressionSetList)
    iterations(xparam) <- 10  
    fit <- xde(xparam, esetList=expressionSetList)
    ##Do more iterations and use a different seed
    firstMcmc(xparam) <- lastMcmc(fit)
    seed(xparam) <- 97814
    fit2 <- xde(xparam, esetList=expressionSetList)

    ##Or
    fit2 <- xde(xparam, esetList=expressionSetList, outputMcmc=fit)
  
## End(Not run)

Options for storing results of the MCMC chains

Description

A numeric vector indicating which chains to write to file and, for those parameters that are written to file, how often the chains should be written to file.

Usage

output(object)
output(object)

Arguments

object

An object of class XdeParameter or XdeMcmc

Details

Replacement methods are only available for objects of class XdeParameter. Accessor methods are available for objects of class XdeParameter and XdeMcmc.

Value

A named numerical vector. The first element (thin) specifies how often to write chains to file. For instance, if output[1]=2 the chains will be written to file every other iteration. Elements 2 - 22 of the vector are indicators for whether to write the write the chains of the Bayesian parameters to file.

Note

Parameters indexed by gene and study (Delta, Phi, Nu, and sigma2) grow very large quickly.

Author(s)

R. Scharpf

Examples

data(xmcmc)
output(xmcmc)
data(xmcmc)
output(xmcmc)

pairs function for high-throughput data

Description

A convenient wrapper for pairs that uses smoothScatter to plot the density of the points and displays the spearman correlation coefficient of the pairwise scatterplots.

Methods

x = "matrix": Typically a matrix of effect size estimates obtained in each study. Rows are genes, columns are studies.
x = "data.frame": Typically a data.frame of effect size estimates obtained in each study. Rows are genes, columns are studies.

Container for XDE parameters

Description

Container for XDE parameters

Objects from the Class

Objects can be created by calls of the form new("Parameters", ...).

Slots

seed:: Object of class "integer" ~~
data:: Object of class "numeric" ~~
phenodata:: Object of class "integer" ~~
G:: Object of class "integer" ~~
Q:: Object of class "integer" ~~
S:: Object of class "integer" ~~
alphaA:: Object of class "numeric" ~~
alphaB:: Object of class "numeric" ~~
betaA:: Object of class "numeric" ~~
betaB:: Object of class "numeric" ~~
pA0:: Object of class "numeric" ~~
pA1:: Object of class "numeric" ~~
pB0:: Object of class "numeric" ~~
pB1:: Object of class "numeric" ~~
nuR:: Object of class "numeric" ~~
nuRho:: Object of class "numeric" ~~
alphaXi:: Object of class "numeric" ~~
betaXi:: Object of class "numeric" ~~
c2Max:: Object of class "numeric" ~~
alphaEta:: Object of class "numeric" ~~
betaEta:: Object of class "numeric" ~~
pOmega0:: Object of class "numeric" ~~
lambdaOmega:: Object of class "numeric" ~~
lambdaKappa:: Object of class "numeric" ~~
gamma2:: Object of class "numeric" ~~
c2:: Object of class "numeric" ~~
tau2Rho:: Object of class "numeric" ~~
tau2R:: Object of class "numeric" ~~
a:: Object of class "numeric" ~~
b:: Object of class "numeric" ~~
l:: Object of class "numeric" ~~
t:: Object of class "numeric" ~~
lambda:: Object of class "numeric" ~~
theta:: Object of class "numeric" ~~
phi:: Object of class "numeric" ~~
sigma2:: Object of class "numeric" ~~
r:: Object of class "numeric" ~~
rho:: Object of class "numeric" ~~
nu:: Object of class "numeric" ~~
delta:: Object of class "numeric" ~~
Delta:: Object of class "numeric" ~~
xi:: Object of class "numeric" ~~

Methods

"[[<-": signature(x = "Parameters"): ...
"[[": signature(x = "Parameters"): ...
"$<-": signature(x = "Parameters"): ...
$: signature(x = "Parameters"): ...
coerce: signature(from = "XdeParameter", to = "Parameters"): ...
show: signature(object = "Parameters"): ...

Examples

showClass("Parameters")
showClass("Parameters")

Accessor and replacement methods for posterior averages of differential expression

Description

Accessor and replacement methods for objects of class XdeMcmc for posterior averages of differential expression

Usage

posteriorAvg(object)
posteriorAvg(object) <- value
posteriorAvg(object)
posteriorAvg(object) <- value

Arguments

`object`	Object of class `XdeMcmc`
`value`	A matrix of dimension G x 3, where G is the number of genes and 3 are different ways of quantifying differential expression in the context of multiple studies (concordant, discordant, or the union).

Value

A matrix of dimension G x 3, where G is the number of genes and 3 are different ways of quantifying differential expression in the context of multiple studies (concordant, discordant, or the union).

Author(s)

Seed for the MCMC

Description

Setting a seed is useful for reproducing MCMC chains

Usage

seed(object)
seed(object) <- value
seed(object)
seed(object) <- value

Arguments

`object`	An object of `XdeParameter` or `XdeMcmc`
`value`	Numeric or integer

Details

The seed stored in the slot of an object of class XdeParameter and an object of class XdeMcmc are useful in different ways. For the XdeParameter class, the seed indicates what seed was used to initialize an MCMC chain. By contrast, an object of class XdeMcmc contains a seed that would be useful for running additional iterations – the seed here is guaranteed to be different from the seed that was used to initiate the MCMC.

Value

An integer

Author(s)

R. Scharpf

Calculate single study estimates of effect size

Description

Calculate single study estimates of effect size for lists of ExpressionSets

Usage

ssStatistic(statistic = c("t", "sam", "z")[1], phenotypeLabel, esetList, ...)
ssStatistic(statistic = c("t", "sam", "z")[1], phenotypeLabel, esetList, ...)

Arguments

`statistic`	Character string indicating Welch t-statistic (t), SAM (sam), or a z-statistic (z)
`phenotypeLabel`	Character string indicating the name of the binary covariate
`esetList`	An object of class `ExpressionSetList`
`...`	Not implemented. Potentially additional arguments to the above methods that are implemented in other packages

Details

This function is a wrapper that provides an estimate of effect size for each study (element) in an ExpressionSetList object.

For Welch t-statistic, this function is a wrapper for mt.teststat in the multtest package.

For SAM, this function is a wrapper for the sam function in the siggenes package.

The "z" statistic is a standardized unbiased estimate of effect size (Hedges and Olkin, 1985) – implementation is in the zScores function in the R package GeneMeta.

See the complete references below.

Value

A matrix: rows are genes and columns are studies

Author(s)

R. Scharpf

References

J.K. Choi, U. Yu, S. Kim, and O.J. Yoo (2003), Combining multiple microarray studies and modeling interstudy variation, Bioinformatics, 19(1) I84-I90.

Y. Ge, S. Dudoit & T. P. Speed (2003), Resampling-based multiple testing for microarray data hypothesis Test 12(1) : 1-44 (with discussions on 44-77).

L. Lusa R. Gentleman, and M. Ruschhaupt, GeneMeta: MetaAnalysis for High Throughput Experiments

L.V. Hedges and I. Olkin, Statistical Methods for Meta-analysis (1985), Academic Press

Tusher, Tibshirani and Chu (2001), Significance analysis of microarrays applied to the ionizing radiation response, PNAS 2001 98: 5116-5121, (Apr 24).

Examples


data(expressionSetList)
if(require(siggenes)){
  sam <- ssStatistic("sam", esetList=expressionSetList, phenotypeLabel="adenoVsquamous")
}  
data(expressionSetList)
if(require(siggenes)){
  sam <- ssStatistic("sam", esetList=expressionSetList, phenotypeLabel="adenoVsquamous")
}

Centers the genes at zero and standardizes the samples to have variance 1

Description

For each study (element) in an ExpressionSetList object, this function centers the genes to have mean zero (rows) and scales the variance of the samples to 1.

Usage

standardizeSamples(object, ...)
standardizeSamples(object, ...)

Arguments

`object`	Object of class `ExpressionSetList`
`...`	Additional arguments not implemented

Value

An object of class ExpressionSetList

Note

Requires genefilter package

Author(s)

R. Scharpf

Center the expression values in a study to zero

Description

Centers each study in a list so that the average expression value of each stuy is zero

Usage

studyCenter(object)
studyCenter(object)

Arguments

object

An object of class ExpressionSetList

Value

An object of class ExpressionSetList

Author(s)

R. Scharpf

Examples

  data(expressionSetList)
  centered <- studyCenter(expressionSetList)
  lapply(centered, function(object) round(mean(exprs(object)), 4))
data(expressionSetList)
  centered <- studyCenter(expressionSetList)
  lapply(centered, function(object) round(mean(exprs(object)), 4))

Useful for changing the look of pairs plots to emphasize concordant or discordant genes

Description

This function can be used to order genes in a matrix by the rank of a statistic and provide different plotting symbols and colors for genes that exceed a certain threshold of the ranking statistic.

Usage

symbolsInteresting(rankingStatistic, percentile = 0.9,
                   colors = c("grey50", "royalblue"),
                   symbols = c(".", "o"), size = c(3, 1),
                   background = c("white", "grey70"))
symbolsInteresting(rankingStatistic, percentile = 0.9,
                   colors = c("grey50", "royalblue"),
                   symbols = c(".", "o"), size = c(3, 1),
                   background = c("white", "grey70"))

Arguments

`rankingStatistic`	Any numerical vector
`percentile`	A percentile of the rankingStatistic – above which a gene would be classified as 'interesting'
`colors`	character string of length 2: a color for genes not exceeding the percentile and a color for genes exceeding the thresold
`symbols`	two plotting symbols (numeric or character): symbol for genes not exceeding percentile and symbol for genes exceeding percentile
`size`	numeric vector of length 2: size of plotting symbol for genes not exceeding percentile and size of plotting symbol for genes exceeding percentile
`background`	character vector of length 2: background color of plotting symbols for gene not exceeding percentile and for genes exceeding the percentile

Value

`order`	the order of the rankingStatistic
`pch`	plotting symbols (same length as rankingStatistic)
`col`	color of plotting symbols (same length as rankingStatistic)
`bg`	background color of plotting symbols (same length as rankingStatistic)
`cex`	size of plotting symbols (same length as rankingStatistic)

Author(s)

R. Scharpf

Examples


  data(expressionSetList)
  data(xmcmc)
  pathToLogFiles <- system.file("logFiles", package="XDE")
  load(file.path(pathToLogFiles, "BES.rda"))
  load(file.path(pathToLogFiles, "postAvg.rda"))
  op.conc <- symbolsInteresting(rankingStatistic=postAvg[, "concordant"])
  graphics:::pairs(BES[op.conc$order, ], pch=op.conc$pch, col=op.conc$col, 
                   bg=op.conc$bg, upper.panel=NULL, cex=op.conc$cex)  

data(expressionSetList)
  data(xmcmc)
  pathToLogFiles <- system.file("logFiles", package="XDE")
  load(file.path(pathToLogFiles, "BES.rda"))
  load(file.path(pathToLogFiles, "postAvg.rda"))
  op.conc <- symbolsInteresting(rankingStatistic=postAvg[, "concordant"])
  graphics:::pairs(BES[op.conc$order, ], pch=op.conc$pch, col=op.conc$col, 
                   bg=op.conc$bg, upper.panel=NULL, cex=op.conc$cex)

How often to write MCMC iterations to file

Description

A value greater than one means that not every MCMC iteration is written to file.

Usage

thin(x, ...)
thin(x, ...)

Arguments

`x`	An object of class `XdeParameter`
`...`	not implemented

Details

thin is an accessor for the first element in the vector returned by the method output.

The replacement method replaces the first element in the output vector.

Value

An integer.

Author(s)

R. Scharpf

Tuning parameters for Metropolis-Hastings proposals

Description

Accessor and replacement methods for tuning the Metropolis-Hastings proposal parameters.

Usage

tuning(object)
tuning(object)

Arguments

object

Object of class XdeParameter

Details

See the XdeParameterClass vignette

Value

A numerical vector

Author(s)

R. Scharpf

Frequency of updating a parameter per MCMC iteration

Description

Accessor and replacement methods for the class XdeParameter are available. Specifying an update of integer N for a Metropolis-Hastings parameter means that N values are proposed for that parameter for each MCMC iteration.

Usage

updates(object)
updates(object)

Arguments

object

An object of class XdeParameter

Details

See the XdeParameterClass vignette

Value

A numerical vector

Author(s)

R. Scharpf

Fit the Bayesian hierarchical model for cross-study differential gene expression

Description

Fits the Bayesian hierarchical model for cross-study differential gene expression.

Usage


  xde(paramsMcmc, esetList, outputMcmc, batchSize=NULL, NCONC=2,
  center=TRUE, ...)

xde(paramsMcmc, esetList, outputMcmc, batchSize=NULL, NCONC=2,
  center=TRUE, ...)

Arguments

`paramsMcmc`	Object of class `XdeParameter`
`esetList`	Object of class `ExpressionSetList`
`outputMcmc`	Object of class `XdeMcmc` (optional)
`batchSize`	Integer or NULL. The number of iterations written to log files before summarizing the chain and then removing. Experimental.
`NCONC`	The number of studies for which a gene must be differentially expressed in the same direction to be considered as concordantly differentially expressed.
`center`	Logical. If TRUE, each study is centered to have mean zero.
`...`	Additional arguments passed to `xdeFit`.

Details

Details for fitting the Bayesian model are discussed elsewhere (see citation below and XdeParameterClass vignette)

If an integer is specified for the batchSize, summary statistics for the log-files are calculated for every batchSize iterations. The log files are then removed and the next iteration will start a new log file. This allows one to do many iterations without creating enormous log files. This is only reasonable to do if one has already assessed convergence.

Value

Object of class XdeMcmc

Note

See the vignettes for XdeParameterClass and XDE.

Author(s)

R. Scharpf

References

R. Scharpf et al., A Bayesian Model for Cross-Study Differential Gene Expression, JASA 2009, p1295–1310.

Examples

## Not run: 
  data(expressionSetList)
  xparam <- new("XdeParameter", phenotypeLabel="adenoVsquamous", esetList=expressionSetList)
  iterations(xparam) <- 10
  fit <- xde(xparam, esetList=expressionSetList)

## End(Not run)
## Not run: 
  data(expressionSetList)
  xparam <- new("XdeParameter", phenotypeLabel="adenoVsquamous", esetList=expressionSetList)
  iterations(xparam) <- 10
  fit <- xde(xparam, esetList=expressionSetList)

## End(Not run)

Class for storing output from the Bayesian model

Description

Stores output, including the last iteration of the MCMC.

Objects from the Class

Objects can be created by calls of the form new("XdeMcmc", studyNames, featureNames, iterations, seed, output, directory, lastMcmc, posteriorAvg, bayesianEffectSize).

Slots

studyNames:: Object of class "character"
featureNames:: Object of class "character"
iterations:: Object of class "numeric"
directory:: Object of class "character"
seed:: Object of class "integer"
output:: Object of class "numeric"
lastMcmc:: Object of class "environment"
posteriorAvg:: Object of class "NULLorMatrix"
bayesianEffectSize:: Object of class "NULLorMatrix"

Methods

$: signature(x = "XdeMcmc")
.standardizedDelta: signature(object = "XdeMcmc")
bayesianEffectSize: signature(object = "XdeMcmc")
bayesianEffectSize<-: signature(object = "XdeMcmc", value = "matrix")
calculatePosteriorAvg: signature(object = "XdeMcmc"): See calculatePosteriorAvg
directory: signature(object = "XdeMcmc")
featureNames: signature(object = "XdeMcmc")
initialize: signature(.Object = "XdeMcmc")
iterations: signature(object = "XdeMcmc")
lastMcmc: signature(object = "XdeMcmc")
nrow: signature(x = "XdeMcmc")
output: signature(object = "XdeMcmc")
plot: signature(x = "XdeMcmc")
posteriorAvg: signature(object = "XdeMcmc")
seed: signature(object = "XdeMcmc")
show: signature(object = "XdeMcmc")
studyNames: signature(object = "XdeMcmc")

Author(s)

R. Scharpf

Examples

##See XDE vignette:
## Not run: 
openVignette(package="XDE")

## End(Not run)
##See XDE vignette:
## Not run: 
openVignette(package="XDE")

## End(Not run)

Container class for storing options of the Bayesian hierarchical model

Description

This class contains initial values for the first iteration of the MCMC, options for saving MCMC chains, options for changing the tuning parameters of the Metropolis-Hastings algorithm, options for changing hyperparameters from their defaults, etc.

Objects from the Class

Objects can be created by calls of the form new("XdeParameter", esetList, updates, tuning, hyperparameters, output, iterations, burnin, seed, randomSeed, genes, studies, firstMcmc, specifiedInitialValues, directory, phenotypeLabel, seed, showIterations, verbose, studyNames, one.delta).

Slots

updates:: Object of class numeric. The frequency of updates for each iteration of the chain.
tuning:: Object of class numeric. Tuning parameters for the Metropolis-Hastings proposals
hyperparameters:: Object of class numeric. Hyperparameters for the Bayesian hierarchical model
output:: Object of class numeric. Indicator for whether to save the MCMC chain to file. If the value is zero, the chain is not saved.
iterations:: Object of class numeric. The total number of MCMC iterations.
burnin:: Object of class logical. If set to FALSE, by default none of the chains will be saved (called for its side-effect of setting the output to zero for each parameter).
notes:: Object of class character.
firstMcmc:: Object of class environment. Values for the first iteration of the MCMC
seed:: Object of class integer. Seed used for simulating random numbers.
showIterations:: Object of class logical. Whether to show the MCMC iteration when fitting the model
specifiedInitialValues:: Object of class logical. If TRUE (the default), the values stored in firstMcmc will be used for the first iteration of the MCMC.
directory:: Object of class character. Specifies where to write the log files
phenotypeLabel:: Object of class character. The name of the binary covariate used for differential expression
verbose:: Object of class logical
studyNames:: Object of class character. Names of the datasets
one.delta:: Logical. If TRUE, a gene is assumed to be differentially in all studies or none of the studies.

Methods

burnin: signature(object = "XdeParameter") logical. See burnin
burnin<-: signature(object = "XdeParameter", value = "logical") logical. See burnin
directory: signature(object = "XdeParameter") character string giving the path or relative path to store log files from the MCMC chain
directory<-: signature(object = "XdeParameter")Path to store log files.
firstMcmc: signature(object = "XdeParameter") See firstMcmc
firstMcmc<-: signature(object = "XdeParameter", value = "environment")
firstMcmc<-: signature(object = "XdeParameter", value = "list")
hyperparameters: signature(object = "XdeParameter") See the XdeParameterClass vignette
hyperparameters<-: signature(object = "XdeParameter") See the XdeParameterClass vignette
initialize: signature(.Object = "XdeParameter") Method for initializing an instance of the class. The default values provided work well in most cases.
iterations: signature(object = "XdeParameter") Accessor for the total number of MCMC iterations to run
iterations<-: signature(object = "XdeParameter", value = numeric) The replacement method is useful for setting a different number of iterations.
iterations<-: signature(object = "XdeParameter", value = "integer")
output: signature(object = "XdeParameter") See also output. This method is also defined for class XdeMcmc
output<-: signature(object = "XdeParameter") See also output
phenotypeLabel: signature(object = "XdeParameter") The name of a binary covariate present in each study
phenotypeLabel<-: signature(object = "XdeParameter", value = "character")
savedIterations: signature(object = "XdeParameter") The number of MCMC iterations written to file. It is the value of the total number of iterations divided by the thinning parameter. See also output
seed: signature(object = "XdeParameter") See seed
seed<-: signature(object = "XdeParameter", value="integer") Replacement method. See also seed.
show: signature(object = "XdeParameter") Produces a short summary of objects that are instances of the XdeParameter class
showIterations: signature(object = "XdeParameter") logical
showIterations<-: signature(object = "XdeParameter")
studyNames: signature(object = "XdeParameter") Names of the high-throughput gene expression studies
studyNames<-: signature(object = "XdeParameter")
thin: signature(x = "XdeParameter") See output and thin
thin<-: signature(x = "XdeParameter", value = numeric) See thin
tuning: signature(object = "XdeParameter") See also tuning
tuning<-: signature(object = "XdeParameter")
updates: signature(object = "XdeParameter") See also updates
updates<-: signature(object = "XdeParameter")

Author(s)

R. Scharpf

References

R. Scharpf

Examples

showClass("XdeParameter")
##See the XdeParameterClass vignette
showClass("XdeParameter")
##See the XdeParameterClass vignette

Object of class XdeMcmc

Description

An object of class XdeMcmc is created by fitting the Bayesian hierarchical model to the expressionSetList example data.

Usage

data(xmcmc)data(xmcmc)

Details

The xmcmc data example was obtained as described in the XDE vignette.

Examples


data(xmcmc)
xmcmc

##ordinarily, one should not need to change the directory in an object
##of class XdeMcmc -- therefore, a replacment method is not defined
pathToLogFiles <- system.file("logFiles", package="XDE")
xmcmc@directory <- pathToLogFiles

##The $ operator can be used to extract chains. For instance, here we
##extract the c2 chain
c2 <- xmcmc$c2
plot.ts(c2)
data(xmcmc)
xmcmc

##ordinarily, one should not need to change the directory in an object
##of class XdeMcmc -- therefore, a replacment method is not defined
pathToLogFiles <- system.file("logFiles", package="XDE")
xmcmc@directory <- pathToLogFiles

##The $ operator can be used to extract chains. For instance, here we
##extract the c2 chain
c2 <- xmcmc$c2
plot.ts(c2)

Alternative cross-study scores of differential expression

Description

Alternative cross-study scores of differential expression

Usage

xsScores(statistic, N)
xsScores(statistic, N)

Arguments

`statistic`	a matrix of study-specific estimates of effect size. Rows are genes and columns are studies.
`N`	numerical vector: the number of samples in each study (the length should be the number of columns in `statistic`)

Value

A matrix of cross-study scores for differential expression ("diffExpressed"), concordant differential expression, and discordant differential expression.

Author(s)

R. Scharpf

References

J.K. Choi, U. Yu, S. Kim, and O.J. Yoo (2003), Combining multiple microarray studies and modeling interstudy variation, Bioinformatics, 19(1) I84-I90.

E. Garrett-Mayer, G. Parmigiani, X. Zhong, L. Cope, and E. Gabrielson (2007), Cross-study validation and combined analysis of gene expression microarray data, Biostatistics, September

R. Scharpf et al., A Bayesian Model for Cross-Study Differential Gene Expression, Technical Report 158, Johns Hopkins University, Department of Biostatistics, 2007

Examples

  data(expressionSetList)
  t <- ssStatistic(statistic="t", phenotypeLabel="adenoVsquamous", esetList=expressionSetList)
  tScores <- xsScores(t, N=nSamples(expressionSetList))
data(expressionSetList)
  t <- ssStatistic(statistic="t", phenotypeLabel="adenoVsquamous", esetList=expressionSetList)
  tScores <- xsScores(t, N=nSamples(expressionSetList))

Option for not modeling Nu

Description

Nu is the average expression value in each study.

Usage

zeroNu(object, ...)
zeroNu(object, ...)

Arguments

`object`	object of class `ExpressionSetList`
`...`	Not implemented

Details

This function should be regarded as experimental.

The nu parameter models the average expression value in each study. Modeling nu allows one to estimate differential expression across studies that may differ in location and scale (as often occurs when multiple platforms are used). The price to pay for modeling nu are additional assumptions (the nu\'s are assumed Gaussian) and a more heavily parameterized model.

The method zeroNu allows one to fit the Bayesian model without estimating nu:

- each gene is centered at zero

- initial values for the first MCMC are chosen on the basis of empirical starting values

- the initial values for a and rho are set to zero.

- the nu, a, gamma2, and rho parameters are not updated during MCMC

Value

object of class XdeParameter

Author(s)

R. Scharpf

References

R. Scharpf et al. (2007), A Bayesian Model for Cross-Study Differential Gene Expression, Technical Report 158, Johns Hopkins University, Department of Biostatistics

Package 'XDE'

Help Index

Indicator for running a MCMC burnin

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Calculate the posterior average for indicators of concordant and discordant differential expression

Description

Usage

Arguments

Details

Value

Author(s)

See Also

Empirical starting values for the MCMC

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Example of ExpressionSetList

Description

Usage

Details

Source

References

Examples

A class for containing a list of ExpressionSets

Description

Objects from the Class

Slots

Extends

Methods

Author(s)

See Also

Examples

Methods for ExpressionSetList

Description

Usage

Arguments

Value

Values for the first MCMC iteration

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Center the expression values for each gene in a study to zero

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Accessor for hyperparameters of the Bayesian model

Description

Usage

Arguments

Details

Value

Author(s)

References

Examples

Number of MCMC iterations

Description

Usage

Arguments

Details

Value

Author(s)

See Also