runAn: Run analysis of an IcaSet object

Description

This function runs the analysis of an ICA decomposition contained in an IcaSet object, according to the parameters entered by the user and contained in a MineICAParams.

Usage

runAn(params, icaSet, keepVar, heatmapCutoff = params["selCutoff"], funClus = c("Mclust", "kmeans"), nbClus, clusterOn = "A", keepComp, keepSamples, adjustBy = c("none", "component", "variable"), typePlot = c("boxplot", "density"), mart = useMart(biomart = "ensembl", dataset = "hsapiens_gene_ensembl"), dbGOstats = c("KEGG", "GO"), ontoGOstats = "BP", condGOstats = TRUE, cutoffGOstats = params["pvalCutoff"], writeGenesByComp = TRUE, writeFeaturesByComp = FALSE, selCutoffWrite = 2.5, runVarAnalysis = TRUE, onlySign = T, runClustering = FALSE, runGOstats = TRUE, plotHist = TRUE, plotHeatmap = TRUE)

Arguments

params

An object of class MineICAParams containing the parameters of the analysis.

icaSet

An object of class IcaSet.

keepVar

The variable labels to be considered, i.e a subset of the annotation variables available in (varLabels(icaSet)).

keepSamples

The samples to be considered, i.e a subset of (sampleNames(icaSet)).

heatmapCutoff

The cutoff (applied to the scaled feature/gene projections contained in S/SByGene) used to select the contributing features/genes.

funClus

The function to be used to cluster the samples, must be one of c("Mclust","kmeans","pam","pamk","hclust","agnes"). Default is "Mclust".

nbClus

The number of clusters to be computed when applying funClus. Can be missing (default) if funClus="Mclust" or funClus="pamk".

keepComp

The indices of the components to be analyzed, must be included in indComp(icaSet). If missing, all components are treated.

adjustBy

The way the p-values of the Wilcoxon and Kruskal-Wallis tests should be corrected for multiple testing: "none" if no p-value correction has to be done, "component" if the p-values have to be corrected by component, "annotation" if the p-values have to be corrected by variable

typePlot

The type of plot used to show distribution of sample-groups contributions, either "density" or "boxplot"

mart

A mart object used for annotation, see function useMart

dbGOstats

The used database to use ('GO' and/or 'KEGG'), default is both.

ontoGOstats

A string specifying the GO ontology to use. Must be one of 'BP', 'CC', or 'MF', see GOHyperGParams. Only used when argument dbGOstats is 'GO'.

condGOstats

A logical indicating whether the calculation should conditioned on the GO structure, see GOHyperGParams.

cutoffGOstats

The p-value threshold used for selecting enriched gene sets, default is params["pvalCutoff"]

writeGenesByComp

If TRUE (default) the gene projections (SByGene(icaSet)) are written in an html file and annotated using biomaRt for each component.

writeFeaturesByComp

If TRUE (default) the feature projections (S(icaSet)) are written in an html file and annotated using biomaRt for each component.

runGOstats

If TRUE the enrichment analysis of the contributing genes is run for each component using package GOstats (default is TRUE).

plotHist

If TRUE the position of the sample annotations within the histograms of the sample contributions are plotted.

plotHeatmap

If TRUE the heatmap of the contributing features/genes are plotted for each component.

runClustering

If TRUE the potential associations between a clustering of the samples (performed according to the components), and the sample annotations, are tested using chi-squared tests.

runVarAnalysis

If TRUE the potential associations between sample contributions (contained in A(icaSet)) are tested using Wilcoxon or Kruskal-Wallis tests.

onlySign

If TRUE (default), only the significant results are plotted in functions

qualVarAnalysis,
  quantVarAnalysis, clusVarAnalysis

, else all plots are done.

selCutoffWrite

The cutoff applied to the absolute feature/gene projection values to select the features/genes that will be annotated using package biomaRt, default is 2.5.

clusterOn

Specifies the matrix used to apply clustering if runClustering=TRUE:

"A":: the clustering is performed in one dimension, on the vector of sample contributions,
"S":: the clustering is performed on the original data restricted to the contributing individuals,
"AS":: the clustering is performed on the matrix formed by the product of the column of A and the row of S.

Value

NULL

Details

This function calls functions of the MineICA package depending on the arguments:

writeProjByComp (if writeGenesByComp=TRUE or writeFeaturesByComp): which writes in html files the description of the features/genes contributing to each component, and their projection values on all the components.
plot_heatmapsOnSel (if plotHeatmap=TRUE): which plots heatmaps of the data restricted to the contributing features/genes of each component.
plotPosAnnotInComp (if plotHist=TRUE): which plots, within the histogram of the sample contribution values of every component, the position of groups of samples formed according to the sample annotations contained in pData(icaSet).
clusterSamplesByComp (if runClustering=TRUE): which clusters the samples according to each component.
clusVarAnalysis (if runClustering=TRUE): which computes the chi-squared test of association between a given clustering of the samples and each annotation level contained in pData(icaSet), and summarizes the results in an HTML file.
runEnrich (if runGOstats=TRUE): which perforns enrichment analysis of the contributing genes of the components using package GOstats.
qualVarAnalysis and quantVarAnalysis (if varAnalysis=TRUE): which tests if the groups of samples formed according to sample annotations contained in pData(icaSet) are differently distributed on the components, in terms of contribution value.

Several directories containing the results of each analysis are created by the function:

ProjByComp:: contains the annotations of the features or genes, one file per component;

varAnalysisOnA:

contains two directories: 'qual/' and 'quant/' which respectively contain the results of the association between components qualitative and quantitative variables;

Heatmaps:

contains the heatmaps (one pdf file per component) of contributing genes by component;

varOnSampleHist:

contains athe histograms of sample contributions superimposed with the histograms of the samples grouped by variable;

cluster2var:

contains the association between a clustering of the samples performed on the mixing matrix A and the variables.

Examples

Run this code

## Not run: 
# 
# ## load an example of IcaSet
# data(icaSetCarbayo)
# ## make sure the 'mart' attribute is correctly defined
# mart(icaSetCarbayo) <- useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
# 
# ## creation of an object of class MineICAParams
# ## here we use a low threshold because 'icaSetCarbayo' is already
# # restricted to the contributing features/genes
# params <- buildMineICAParams(resPath="~/resMineICACarbayotestRunAn/", selCutoff=2, pvalCutoff=0.05)
# require(hgu133a.db)
# 
# runAn(params=params, icaSet=icaSetCarbayo)
# ## End(Not run)