Package 'bgx'

Analyse BGX output.

Description

Functions for plotting expression densities, differential expression densities, histogram of proportion of differentially expressed genes, etc.

Usage

plotExpressionDensity(bgxOutput, gene=NULL, normalize=c("none","mean","loess"),...)
  plotDEDensity(bgxOutput, gene=NULL, conditions=c(1,2), normalize=c("none","mean","loess"), normgenes=c(1:length(bgxOutput[["geneNames"]])), ...)
  plotDEHistogram(bgxOutput, conditions=c(1,2), normalize=c("none","mean","loess"), normgenes=c(1:length(bgxOutput[["geneNames"]])), df=floor(1.8 * log10(length(bgxOutput[["geneNames"]]))))
  rankByDE(bgxOutput, conditions=c(1,2),normalize=c("none", "mean", "loess"), normgenes=c(1:length(bgxOutput[["geneNames"]])), absolute=TRUE)
  plotDiffRank(bgxOutput, conditions=c(1,2),normalize=c("none", "mean", "loess"), normgenes=c(1:length(bgxOutput[["geneNames"]])), ymax=NULL)

Arguments

bgxOutput	A list obtained from running readOutput.bgx on a BGX output directory.
gene	Which gene to analyse. This can either be an index or a name.
conditions	Indices of conditions to compare.
normalize	"none": do not normalise posterior distributions of mu. "mean": normalise by scaling posterior distributions of mu for conditions > 1 to have the same mean as the posterior distribution of mu for condition 1. "loess": same as "mean" but use loess normalisation.
normgenes	Which genes to use for loess normalisation. By default, use all genes.
df	Residual degrees of freedom. Decrease to 6 if the histogram fit goes haywire.
absolute	Rank genes by absolute differential expression.
ymax	Specify upper limit of y axis.
...	Parameters to pass to density function (where applicable).

Details

plotExpressionDensity plots gene expression distributions under various conditions for the specified gene.

plotDEDensity plots the differential expression distribution between two conditions for a given gene.

plotDEHistogram plots a histogram of differential expression between two conditions and estimates the number of up and down regulated differentially expressed genes.

rankByDE ranks genes by differential expression and returns ordering and corresponding DE values in a matrix.

plotDiffRank plots 2.5-97.5% confidence intervals for ranked differential expression estimates.

Value

None, except plotDERank, which returns a matrix of genes ranked by differential expression.

Author(s)

Ernest Turro

See Also

bgx, standalone.bgx, readOutput.bgx, plotExpressionDensity, plotDEDensity, plotDEHistogram

---

Fully Bayesian integrated approach to the analysis of Affymetrix GeneChip data

Description

'bgx' estimates Bayesian Gene eXpression (BGX) measures from an AffyBatch object.

'standalone.bgx' creates various files needed by the bgx standalone binary and places them in a directory. One of these files is 'infile.txt'. In order to run standalone BGX, compile it and run 'bgx <path\_to\_infile.txt>' from the command line.

Usage

bgx(aData, samplesets = NULL, genes = NULL, genesToWatch = NULL,
  burnin = 8192, iter = 16384, output = c("minimal","trace","all"), 
  probeAff = TRUE, probecat_threshold = 100, adaptive = TRUE, rundir = ".")

standalone.bgx(aData, samplesets = NULL, genes = NULL, genesToWatch = NULL,
  burnin = 8192, iter = 16384, output = c("minimal", "trace", "all"),
  probeAff = TRUE, probecat_threshold = 100,
  adaptive = TRUE, batch_size = 50, optimalAR = 0.44, inputdir = "input")

Arguments

aData	An AffyBatch object.
samplesets	A numeric vector specifying which condition each array belongs to. E.g. if samplesets=c(2,2), then the first two replicates belong to one condition and the last two replicates belong to another condition. If NULL, each array is assumed to belong to a different condition. If the aData object contains information about the experiment design in its phenoData slot, this argument is not required.
genes	A numeric vector specifying which genes to analyse. If NULL, all genes are analysed.
genesToWatch	A numeric vector specifying which genes to monitor closely amongst those chosen to be analysed (see below for details).
burnin	Number of burn-in iterations.
iter	Number of post burn-in iterations.
output	One of "minimal", "trace" or "all". See below for details.
probeAff	Stratify the mean (lambda) of the cross-hybridisation parameter (H) by categories according to probe-level sequence information.
probecat_threshold	Minimum amount of probes per probe affinity category.
adaptive	Adapt the variance of the proposals for Metropolis Hastings objects (that is: S, H, Lambda, Eta, Sigma and Mu).
batch_size	Size of batches for calculating acceptance ratios and updating jumps.
optimalAR	Optimal acceptance ratio.
rundir	The directory in which to save the output runs.
inputdir	The name of the directory in which to place the input files for the standalone binary.

Details

• genesToWatchSpecify the subset of genes for which thinned samples from the full posterior distributions of log(S+1) (x) and log(H+1) (y) are collected.

• outputOutput the following to disk:

  • "minimal"The gene expression measure (muave), thinned samples from the full posterior distributions of mu (mu.[1..c]), where 'c' is the number of conditions, the integrated autocorrelation time (IACT) and the Markov chain Monte Carlo Standard Error (MCSE) for each gene under each condition. Note that the IACT and MCSE are calculated from the thinned samples of mu.

  • "trace"The same as "minimal" plus thinned samples from the full posterior distributions of sigma2 (sigma2.[1..c]), lambda (lambda.[1..s]), eta2 (eta2), phi (phi) and tau2 (tau2), where 's' is the number of samples. If there are probes with unknown sequences, output a thinned trace of their categorisation.

  • "all"The same as "trace" plus acceptance ratios for S (sacc), H (hacc), mu (muacc), sigma (sigmaacc), eta (etaacc) and lambda (lambdasacc).

Value

'bgx' returns an ExpressionSet object containing gene expression information for each gene under each condition (not each replicate).

'standalone.bgx' returns the path to the BGX input files.

Note

The bgx() method and the bgx standalone binary create a directory in the working directory called 'run.x' (x:1,2,3,...), wherein files are placed for further detailed analysis.

Author(s)

Ernest Turro

References

Turro, E., Bochkina, N., Hein, A., Richardson, S. (2007) BGX: a Bioconductor package for the Bayesian integrated analysis of Affymetrix GeneChips. BMC Bioinformatics 2007, 8:439.

Hein, A., Richardson, S. (2006) A powerful method for detecting differentially expressed genes from GeneChip arrays that does not require replicates. BMC Bioinformatics 2006, 7:353.

Hein, A., Richardson, S., Causton, H., Ambler, G., Green., P. (2005) BGX: a fully Bayesian integrated approach to the analysis of Affymetrix GeneChip data. Biostatistics, 6, 3, pp. 349-373.

Hekstra, D., Taussig, A. R., Magnasco, M., and Naef, F. (2003) Absolute mRNA concentrations from sequence-specific calibration of oligonucleotide array. Nucleic Acids Research, 31. 1962-1968.

G.O. Roberts, J.S. Rosenthal (September, 2006) Examples of Adaptive MCMC.

Examples

# This example requires the 'affydata' and 'hgu95av2cdf' packages 
  if(require(affydata) && require(hgu95av2cdf)) {
    data(Dilution)
    eset <- bgx(Dilution, samplesets=c(2,2), probeAff=FALSE, burnin=4096, iter=8192,
      genes=c(12500:12599), output="all", rundir=file.path(tempdir()))
  }

---

Read in the output from a BGX run.

Description

readOutput.bgx reads in output from BGX which can then be fed into BGX analysis functions.

Usage

readOutput.bgx(...)

Arguments

...	Paths of BGX output directories. If you specify more than one path, then the runs will be combined such that each condition from each run is treated as a different different from all the others.

Details

See bgx for more details.

Value

A list containing data from the BGX output.

Author(s)

Ernest Turro

See Also

bgx, standalone.bgx, plotExpressionDensity, plotDEDensity, plotDEHistogram

---