##' Object of class \code{\link{IcaSet}} containing an ICA decomposition calculated by the FastICA algorithm
##' (through matlab function "icasso") on
##' bladder cancer expression data measured on HG-U133A Affymetrix microarrays.
##' The original expression data were normalized with MAS5 by the authors of the paper
##' followed by log2-transformation.
##' ICA was run on the dataset restricted to the 10000 most variable probe sets (based on IQR values).
##' 10 components were computed.
##' Only probe sets/genes having an absolute projection higher than 3 are stored in this object.
##' @title IcaSet-object containing a FastICA decomposition of gene expression microarrray-based data of bladder cancer samples.
##' @name icaSetCarbayo
##' @docType data
##' @author Anne Biton
##' @references \url{http://jco.ascopubs.org/content/24/5/778/suppl/DC1}
##' @keywords datasets
NULL
##' readA
##'
##' This function reads and annotates matrix A.
##'
##' The matrix dat must be the one on which the matrix A was calculated.
##' It is assumed that the number of components is lower than the number of
##' samples, the matrix will be transposed to have dimension 'samples x components' according to this assumption. If \code{annot} is FALSE, colnames of dat are used to annotate
##' rownames of A.
##' @title read A
##' @param Afile The file which contains the matrix of sample contributions. It
##' must be a txt file where the separator is \code{white space}, that is one or
##' more spaces, tabs, newlines or carriage returns
##' @param datfile The file which contains the matrix (of dimension features x
##' samples) based on which the matrix A was calculated
##' @param dat The data based on which the matrix A was calculated (features x
##' samples)
##' @param annot TRUE (default) if the Afile contains rownames of matrix A,
##' FALSE if the rownames has to be extracted from dat
##' @return This function returns a matrix of dimension samples x components
##' with rownames filled with sample IDs.
##' @keywords internal
##' @author Anne Biton
readA <- function (
Afile
,datfile
,dat
,annot = TRUE
)
{
message("..Reading matrix A.. ")
A <- read.table(Afile, header = if (!annot) TRUE else FALSE) #sep = sep,
nr <- nrow(A)
nc <- ncol(A)
if (nr < nc)
A <- t(A)
A <- as.data.frame(A, check.names = FALSE, stringsAsFactors = FALSE)
if (annot) {
if (missing(dat)) {
if (missing(datfile))
stop("Either the expression matrix or the corresponding file must be provided.")
dat <- read.table(datfile, header = TRUE, nrows = 1, stringsAsFactors= FALSE, check.names = FALSE)
if (ncol(dat) != nrow(A))
stop("The number of rows of A must equal the number of columns of dat.")
}
rownames(A) = colnames(dat)
}
gc()
return(A)
}
##' This function reads and annotates matrix S.
##'
##' The matrix dat must be the one on which the matrix S was calculated.
##' It is assumed that the number of components is lower than the number of features, the matrix will be transposed to have dimension 'features x components' according to this assumption.
##' If \code{annot} is FALSE, rownames of dat are used to annotate rownames of S.
##' @title read S
##' @param Sfile The file which contains the matrix of feature projections. It
##' must be a txt file where the separator is \code{white space}, that is one or
##' more spaces, tabs, newlines or carriage returns.
##' @param datfile The file which contains the matrix (of dimension features x
##' samples) based on which the matrix S was calculated. It must be a txt file
##' where the separator is \code{white space}, that is one or more spaces, tabs,
##' newlines or carriage returns.
##' @param dat The data based on which the matrix A was calculated (features x
##' samples)
##' @param annot TRUE (default) if the Afile contains rownames of matrix A,
##' FALSE if the rownames has to be extracted from dat
##' @return This function returns a matrix of dimension features x components
##' with rownames filled with feature IDs.
##' @keywords internal
##' @author Anne Biton
readS <- function (Sfile,
datfile,
dat,
annot = TRUE)
{
message("..Reading matrix S..")
S <- read.table(Sfile, header = if (!annot) TRUE else FALSE)# sep = sep,
nr <- nrow(S)
nc <- ncol(S)
if (nr < nc)
S <- t(S)
S <- as.data.frame(S, check.names = FALSE, stringsAsFactors = FALSE)
if (annot) {
message("...Put feature labels on S matrix...")
if (missing(dat)) {
if (missing(datfile))
stop("Either the data matrix or the corresponding file must be provided.")
dat <- read.table(datfile, header = TRUE, check.names = FALSE, stringsAsFactors = FALSE)
}
rownames(S) <- rownames(dat)
}
gc()
return(S)
}
##
## Builds a subset of an IcaSet object
##
## \code{keepSamples} must be available in \code{sampleNames(icaSet)}.
## \code{keepComp} must be available in indComp(icaSet), i.e if \code{icaSet} has been previously restricted to a few components by subIcaSet, \code{keepComp} must refer to the index of the components in the original icaSet object.
## @title subset an \code{\link{IcaSet}} object
## @param icaSet An IcaSet object
## @param keepSamples The sample IDs to be selected, must be included in \code{sampleNames(icaSet)}. If missing, all samples are kept.
## @param keepComp The component indices to be selected, must included in \code{indComp(icaSet)}. If missing, all components are kept.
## @return This function returns an \code{IcaSet} object.
## @author Anne Biton
## @keywords internal
## @seealso \code{\link{IcaSet-class}}
## @examples
## ## load an example of IcaSet
## data(icaSetCarbayo)
##
## #keep a subset of samples (samples of stage T2, T3, T4)
## invSamples <- rownames(subset(pData(icaSetCarbayo),stage %in% c("T2","T3","T4")))
## icaSetCarbayo_subSamples <- subIcaSet(icaSetCarbayo, keepSamples=invSamples)
##
## #keep a subset of components (components of index 5:9 and 11,12)
## icaSetCarbayo_subComp <- subIcaSet(icaSetCarbayo, keepComp=c(5:9,11,12))
##
subIcaSet <- function(
icaSet
,keepSamples
,keepComp
) {
if (!missing(keepSamples)) {
diffSamples <- setdiff(keepSamples,sampleNames(icaSet))
if (length(diffSamples) == length(keepSamples))
stop("The sample ids provided in 'keepSamples' are not available in 'icaSet'.")
else if (length(diffSamples) > 0)
warning(paste("The samples:",paste(diffSamples,collapse=", "),"are not available in icaSet."))
keepSamples <- intersect(keepSamples,sampleNames(icaSet))
icaSet@A <- A(icaSet)[keepSamples,]
varLabelsor <- varLabels(icaSet)
pData(icaSet) <- pData(icaSet)[keepSamples,,drop=FALSE]
varLabels(icaSet) <- varLabelsor
pData(protocolData(icaSet))<- pData(protocolData(icaSet))[keepSamples,,drop=FALSE]
assayDataElement(icaSet,"dat") <- assayDataElement(icaSet,"dat")[,keepSamples,drop=FALSE]
icaSet@datByGene <- datByGene(icaSet)[,keepSamples,drop=FALSE]
refSamples(icaSet) <- intersect(refSamples(icaSet),keepSamples)
}
if (!missing(keepComp)) {
if (!is.numeric(keepComp))
stop ("keepComp must be numeric")
nbcomp <- nbComp(icaSet)
if (sum(sort(keepComp) %in% indComp(icaSet)) != nbcomp) {
diffcomp <- setdiff(keepComp,indComp(icaSet))
if (length(diffcomp) == length(keepComp))
stop("The component numbers provided in 'keepComp' are not included in 'indComp(icaSet)'.")
else if (length(diffcomp) > 0)
warning(paste("The component(s):",paste(diffcomp,collapse=", ",sep=""),"are not available in 'icaSet'."))
keepComp <- intersect(keepComp,indComp(icaSet))
keepCompor <- keepComp
keepComp <- match(keepComp, indComp(icaSet))
icaSet@compNames <- compNames(icaSet)[keepComp]
if (length(witGenes(icaSet))>0)
icaSet@witGenes <- witGenes(icaSet)[keepComp]
icaSet@A <- A(icaSet)[,keepComp, drop=FALSE]
icaSet@S <- S(icaSet)[,keepComp, drop=FALSE]
icaSet@SByGene <- SByGene(icaSet)[,keepComp, drop=FALSE]
indComp(icaSet) <- keepCompor
}
}
validObject(icaSet)
return(icaSet)
}
##' Extract projection values of a given set of IDs on a subset of components.
##'
##'
##' @title Extract projection values
##' @param icaSet An object of class \code{\link{IcaSet}}
##' @param ids feature or gene IDs
##' @param keepComp Index of the components to be conserved, must be in \code{indComp(icaSet)}
##' @param level The level of projections to be extracted, either \code{"features"} or \code{"genes"}
##' @return A vector or a list of projection values
##' @author Anne Biton
##' @export
##' @examples
##'
##' ## load an example of IcaSet
##' data(icaSetCarbayo)
##'
##' ##get the projection of your favorite proliferation genes
##' #on all components
##' getProj(icaSetCarbayo, ids=c("TOP2A","CDK1","CDC20"), level="genes")
##'
##' #on some components
##' getProj(icaSetCarbayo, ids=c("TOP2A","CDK1","CDC20"),
##' keepComp=c(1,6,9,12),level="genes")
##'
##' ##get the gene projection values on the sixth component
##' getProj(icaSetCarbayo, keepComp=6,level="genes")
##'
getProj <- function(icaSet,ids,keepComp, level = c("features","genes")) {
level <- match.arg(tolower(level), choices=c("features","genes"))
switch(level,
features={ data=S(icaSet)},
genes={data=SByGene(icaSet)}
)
if (!missing(ids)) {
interG <- intersect(rownames(data),ids)
if (length(interG) == 0)
stop("The given ids are not available in icaSet")
else if (length(interG) < length(ids)) {
diff <- setdiff(ids,rownames(data))
warning(paste("The IDs",paste(diff,collapse = ", "),"are not avaible in icaSet"))
}
}
else interG <- rownames(data)
if (!missing(keepComp)) {
diffcomp <- setdiff(keepComp,indComp(icaSet))
if (length(diffcomp) == length(keepComp))
stop("The component numbers provided in 'keepComp' are not included in 'indComp(icaSet)'.")
else if (length(diffcomp) > 0)
warning(paste("The component(s):",paste(diffcomp,collapse=", ",sep=""),"are not available in 'icaSet'."))
keepComp <- intersect(keepComp,indComp(icaSet))
}
else
keepComp <- indComp(icaSet)
out <- data[interG,keepComp]
if (length(keepComp)==1)
names(out) <- interG
return(out)
}
##' This function annotates the features of an \code{\link{IcaSet}} object and fills its attributes \code{SByGene} and \code{datByGene}.
##'
##' When attribute \code{annotation} of \code{icaSet} is not specified (of length \code{0}), \code{biomaRt} is used to annotate the features through function \code{\link{annotFeaturesWithBiomaRt}}.
##'
##' When specified, attribute \code{annotation} of argument \code{icaSet} must be an annotation package and will be used to annotate the \code{featureNames} of \code{icaSet}. In addition, the attribute \code{typeID} (a vector) of argument \code{icaSet} must contain a valid element \code{geneID_annotation} that determines the object of the package to be used for the annotation, see \code{\link{IcaSet}}.
##'
##' When argument \code{annot} is TRUE, this function fills the attributes \code{SByGene} and \code{datByGene} of \code{icaSet}. When several feature IDs are available for a same gene ID, the median value of the corresponding features IDs is attributed to the gene (the median of the projection values is used for attribute \code{SByGene}, and the median of the expression values is used for attribute \code{datByGene}).
##'
##' When attribute \code{chipManu} of the argument \code{icaSet} is "illumina", the features are first converted into nuID using the package 'lumi*Mapping' and then annotated into genes.
##' In that case, features can only be annotated in ENTREZID or SYMBOL. It means that \code{typeID(icaSet)['geneID_annotation']} must be either 'ENTREZID' or 'SYMBOL'. You will need to annotate yourself the IcaSet object if you want to use different IDs.
##'
##' @title Features annotation of an object of class IcaSet.
##' @param icaSet An object of class \code{\link{IcaSet}} to be annotated, must contain a valid \code{annotation} attribute.
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the analysis.
##' @param annot TRUE (default) if the IcaSet object must indeed be annotated
##' @return The modified argument \code{icaSet}, with filled attributes \code{SByGene} and \code{datByGene}.
##' @seealso \code{\link{annotFeaturesComp}}
##' @author Anne Biton
##' @export
##' @examples
##' #load data
##' data(icaSetCarbayo)
##' require(hgu133a.db)
##'
##' # run annotation of the features into gene Symbols as specified in 'typeID(icaSetCarbayo)["geneID_annotation"]',
##' # using package hgu133a.db as defined in 'annotation(icaSetMainz)'
##' icaSetCarbayo <- annotInGene(icaSet=icaSetCarbayo, params=buildMineICAParams())
##'
##' \dontrun{
##' #load data
##' library(breastCancerMAINZ)
##' data(mainz)
##' #run ICA
##' resJade <- runICA(X=exprs(mainz), nbComp=5, method = "JADE", maxit=10000)
##'
##' #build params
##' params <- buildMineICAParams(resPath="mainz/")
##'
##' #build a new IcaSet object, omitting annotation of the features (runAnnot=FALSE)
##' #but specifying the element "geneID_annotation" of argument 'typeID'
##' icaSetMainz <- buildIcaSet(params=params, A=data.frame(resJade$A), S=data.frame(resJade$S),
##' dat=exprs(mainz), pData=pData(mainz),
##' annotation="hgu133a.db", typeID= c(geneID_annotation = "SYMBOL",
##' geneID_biomart = "hgnc_symbol", featureID_biomart = "affy_hg_u133a"),
##' chipManu = "affymetrix", runAnnot=FALSE,
##' mart=useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl"))
##'
##' #Attributes SByGene is empty and attribute datByGene refers to assayData
##' SByGene(icaSetMainz)
##' head(datByGene(icaSetMainz))
##'
##' # run annotation of the features into gene Symbols as specified in 'typeID(icaSetMainz)["geneID_annotation"]',
##' # using package hgu133a.db as defined in 'annotation(icaSetMainz)'
##' icaSetMainz <- annotInGene(icaSet=icaSetMainz, params=params)
##' }
annotInGene <- function(icaSet,
params,
annot = TRUE
) {
chip <- annotation(icaSet)
if (annot) {
if (length(chip)==0 || (chip == ""))
message("No annotation package is available, biomaRt is used for the annotation.")
else {
if (!("geneID_annotation" %in% names(typeID(icaSet))))
stop(paste("Since you want to annotate the features using ", chip,", typeID attribute of IcaSet object must contain a 'geneID_annotation' element which determines the object from the package you want to use.", sep = ""))
else {
listIDs <- gsub(x=ls(paste("package:",chip,sep="")), pattern = gsub(chip,pattern=".db",replacement = ""), replacement = "")
if (!(typeID(icaSet)["geneID_annotation"] %in% listIDs))
stop(paste("The element 'geneID_annotation' of attribute 'typeID' of object IcaSet is not available in annotation package,",chip))
library(chip, character.only = TRUE)
}
}
icaSet <- annotFeaturesComp(icaSet = icaSet, params=params, type = toupper(typeID(icaSet)["geneID_annotation"]))
}
return(icaSet)
}
##' This function annotates a set of features
##'
##' @title Annotation of features using an annotation package
##' @param features Feature IDs to be annotated
##' @param type The object from the package used to annotate the features, must be available in \code{ls("package:package_name")}
##' @param annotation An annotation package
##' @return A vector of gene/object IDs indexed by the feature IDs.
##' @author Anne Biton
##' @export
##' @examples
##' library(hgu133a.db)
##' annotFeatures(features = c("1007_s_at", "1053_at", "117_at", "121_at", "1255_g_at"),
##' type="SYMBOL", annotation="hgu133a.db")
##'
annotFeatures <- function(features,
type,
annotation
) {
pack.annot <- eval(as.name(paste(gsub(".db", "", annotation), type, sep = "")))
probeToGene = unlist(AnnotationDbi::mget(features, pack.annot, ifnotfound = NA))
names(probeToGene) = features
probeToGene = na.omit(probeToGene)
message(paste("The number of probe sets not associated with a",type,"ID is",length(features)-length(probeToGene)))
message(paste("The remaining number of probe sets is",length(probeToGene)))
return(probeToGene)
}
## ## annotation of an IcaSet object (icaSetKim) containing data from an illumina microarray
## # chipManu must be available and equal to "illumina"
## chipManu(icaSetKim)
## # annotation must be available and equal to "lumi*All.db" where * is the organism, in this case 'Human'.
## annotation(icaSetKim)
## icaSetKim_annot <- annotFeaturesComp(icaSet=icaSetKim, params=params, type="SYMBOL")
##' ##' This function annotates the features of an object of class \code{\link{IcaSet}}, and fills its attributes \code{SByGene} and \code{datByGene}.
##'
##' This function is called by function \code{\link{annotInGene}} which will check the validity of the attributes \code{annotation, typeID, chipManu} and eventually \code{chipVersion} of \code{icaSet}.
##' If available, the attribute \code{annotation} of argument \code{icaSet} must be an annotation package and will be used to annotate the \code{featureNames} of \code{icaSet}.
##' If attribute \code{annotation} of argument \code{icaSet} is not available (of length 0), \code{biomaRt} is used to annotate the features.
##'
##' This function fills the attributes \code{SByGene} and \code{datByGene} of the argument \code{icaSet}. When several feature IDs are available for a same gene ID, the median value of the corresponding features IDs is attributed to the gene (the median of projection values is used for attribute \code{SByGene}, and the median of expression values is used for attribute \code{datByGene}).
##'
##' When attribute \code{chipManu} of the argument \code{icaSet} is "illumina", the features are first converted into nuID using the package 'lumi*Mapping' and then annotated into genes.
##' In that case, features can only be annotated in ENTREZID or SYMBOL. It means that \code{typeID(icaSet)['geneID_annotation']} must be either 'ENTREZID' or 'SYMBOL'. You will need to annotate yourself the \code{\link{IcaSet}} object if you want to use different IDs.
##'
##'
##' @title Features annotation
##' @param icaSet An object of class \code{\link{IcaSet}} whose features have to be annotated. The attribute \code{annotation} of this object contains the annotation package to be used.
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the analysis.
##' @param type The ID of the object of the annotation package to be used for the annotation, must be available in \code{ls("package:package_name")}
##' @param featureId The type of the feature IDs, in the \code{biomaRt} way (type \code{listFilters(mart)} to choose one). Used when \code{annotation(icaSet)} is of length 0.
##' @param geneId The type of the gene IDs, in the \code{biomaRt} way (type \code{listAttributes(mart)} to choose one). Used when \code{annotation(icaSet)} is of length 0.
##' @return This function returns the argument \code{icaSet} with attributes
##' \code{SByGene} and \code{datByGene} filled.
##' @author Anne Biton
##' @export
##' @seealso \code{\link{annotFeatures}}, \code{\link{annotFeaturesWithBiomaRt}}, \code{\link{annotInGene}}
##' @examples
##'
##' ## load an example of IcaSet
##' data(icaSetCarbayo)
##' params <- buildMineICAParams()
##' require(hgu133a.db)
##' ####===================================================
##' ## Use of annotation package contained in annotation(icaSet)
##' ####====================================================
##' ## annotation in SYMBOL
##' icaSetCarbayo_annot <- annotFeaturesComp(icaSet=icaSetCarbayo, params=params, type="SYMBOL")
##' # arg 'type' is optional since the function uses contents of typeID(icaSet) as the defaults,
##' # it is specified in these examples for pedagogy views
##'
##' ## annotation in Entrez Gene
##' icaSetCarbayo_annot <- annotFeaturesComp(icaSet=icaSetCarbayo, params=params, type="ENTREZID")
##'
##' \dontrun{
##' ####===================================================
##' ## Use of biomaRt, when annotation(icaSet) is of length 0
##' ####====================================================
##' ## empty attribute 'annotation' of the IcaSet object
##' # when this attribute is not specified, biomaRt is used for annotation
##' annotation(icaSetCarbayo) <- character()
##'
##' # make sure the mart attribute is correctly defined
##' mart(icaSetCarbayo) <- useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
##'
##' ## make sure elements "featureID_biomaRt" and "geneID_biomaRt" of typeID(icaSet) are correctly filled
##' # they will be used by function 'annotFeaturesComp' through biomaRt to query the database
##' typeID(icaSetCarbayo)
##'
##' ## run annotation of HG-U133A probe set IDs into Gene Symbols using biomaRt
##' icaSetCarbayo_annot <- annotFeaturesComp(icaSet=icaSetCarbayo, params=params)
##'
##' }
annotFeaturesComp <- function(icaSet,
params,
type = toupper(typeID(icaSet)["geneID_annotation"]),
featureId = typeID(icaSet)["featureID_biomart"],
geneId = typeID(icaSet)["geneID_biomart"]
) {
if (nrow(S(icaSet))==0)
stop("Attribute 'S' is missing in icaSet")
if (length(chipManu(icaSet))>0)
manu <- match.arg(tolower(chipManu(icaSet)), c("illumina","affymetrix","agilent"))
else
manu <- ""
## if annotation package not provided, annotation with biomaRt
if (length(annotation(icaSet))==0 || annotation(icaSet)=="") {
message("..Features annotation with biomaRt..","\n")
if (mart(icaSet)@dataset == "")
stop("Please fill the mart attribute of the 'icaSet' object using the function 'useMart'.")
message("...Annotation of ",featureId," into ",geneId,"\n")
probe2gene <- annotFeaturesWithBiomaRt(features=featureNames(icaSet), featureId=featureId, geneId=geneId, mart=mart(icaSet))
}
else {
message("..Features annotation with package ",annotation(icaSet),"..","\n")
message("...Annotation into ",typeID(icaSet)["geneID_annotation"],"..","\n")
if (manu == "illumina") {
if ((!(toupper(type) %in% c("ENTREZID", "SYMBOL") )))
stop("For illumina, features can only be annotated in ENTREZID or SYMBOL. You will need to annotate yourself the IcaSet object if you want to use different IDs.")
ilmn_features = unique(featureNames(icaSet))
message("....Mapping features<->nuID.... \n")
if (length(agrep(organism(icaSet),c("Homo Sapiens", "HomoSapiens")))>0)
spec <- "Human"
else
spec <- c("Human","Rat","Mouse")[agrep(organism(icaSet), c("Human","Rat","Mouse"))]
if (length(spec) == 0)
spec <- "Unknown"
message("....Loading illumina mapping package needed for annotation....\n")
eval(parse(text=c(Human = "library('lumiHumanIDMapping');library('lumiHumanAll.db')", Rat = "library('lumiRatIDMapping');library('lumiRatAll.db')", Mouse = "library('lumiMouseIDMapping');library('lumiMouseAll.db')")[spec]))
### Mapping Illumina features <-> nuIDs
chipInfo <- lumi:::getChipInfo(ilmn_features, species = spec, lib.mapping = c(Human = "lumiHumanIDMapping", Rat = "lumiRatIDMapping", Mouse = "lumiMouseIDMapping")[spec], chipVersion = if (length(chipVersion(icaSet))>0 | chipVersion(icaSet) != "") chipVersion(icaSet) else NULL, idMapping = TRUE, verbose = FALSE)
# nuID to illumina probe
mapping <- chipInfo$idMapping
newId <- mapping[, "nuID"]
names(newId) = rownames(mapping)
## restrict to the common genes
newId = newId[intersect(names(newId),featureNames(icaSet))]
message("......mapping nuID<->genes......\n")
## Mapping nuIDs <-> genes
## Now I need to retrieve the annotations of the illumina features thanks to their nuIDs.
## Using lumiHumanAll.db package
if (type == "ENTREZID") {
nuid2genes = getEG(newId, "lumiHumanAll.db")
}
else if (type == "SYMBOL") {
nuid2genes = getSYMBOL(newId, "lumiHumanAll.db")
}
naInd <- which(is.na(nuid2genes))
nuid2genes <- nuid2genes[which(!is.na(nuid2genes))]
message(paste("Number of features = ",length(ilmn_features), "\n"))
message(paste("Number of features with no associated gene id = ",length(naInd), "\n"))
message(paste("Number of features with an associated gene id = ",length(nuid2genes), "\n"))
message(paste("Number of gene ids = ",length(unique(nuid2genes)), "\n"))
message(".......mapping illumina features<->genes....... \n")
### gene <-> illumina probe
# nuid to probe
probe2nuid = newId
#names(probe2nuid) = names(newId)
probe2gene = nuid2genes[probe2nuid] #long
names(probe2gene) = names(probe2nuid)
}
else {
# all features
allfeatures = featureNames(icaSet)
# probe -> gene
probe2gene = annotFeatures(allfeatures, type = type, annotation = annotation(icaSet))
noms = names(probe2gene)
probe2gene = as.character(probe2gene) #seul moyen de le transformer en vecteur
names(probe2gene) = noms
probe2gene = probe2gene[intersect(names(probe2gene),featureNames(icaSet))]
}
}
probe2gene <- probe2gene[!is.na(probe2gene)]
allGenes = unique(unlist(probe2gene))
## List : Gene -> c(features)
gene2probe <- lapply(allGenes,
function(geneId, probe2gene) {
selecProbeGenes = probe2gene[probe2gene %in% geneId]
return(names(selecProbeGenes))
}
, probe2gene = probe2gene
)
names(gene2probe) <- allGenes
### Is there only one probe associated to each gene (use of BrainArray for example)?
gene2nbfeatures <- unlist(llply(gene2probe,length))
if (all(gene2nbfeatures == 1)) {
oneProbeByGene = TRUE
message("Each gene is annotated by only one probe set. \n")
gene2probe = unlist(gene2probe)
}
else
oneProbeByGene = FALSE
names(gene2nbfeatures) <- names(gene2probe) <- allGenes
genes2oneprobe = names(gene2nbfeatures[gene2nbfeatures == 1])
dat <- assayDataElement(icaSet,"dat")
dat_genes = dat[unlist(gene2probe[genes2oneprobe]), ]
rownames(dat_genes) = as.character(genes2oneprobe)
#### Gene -> median features
if (!oneProbeByGene) {
genesmanyfeatures = names(gene2nbfeatures[gene2nbfeatures > 1])
message(paste ("The number of genes annotated by several features is", length(genesmanyfeatures[!is.na(genesmanyfeatures)]), "\n"))
featuresmanygenes = unlist(gene2probe[genesmanyfeatures])
gene2probe_manyfeatures = gene2probe[genesmanyfeatures]
### Gene -> Median of the probe sets projection
genesComp = llply(Slist(icaSet),
function (probe2proj, probe2gene, genesmanyfeatures, featuresmanygenes, gene2probe_manyfeatures) {
orOrd <- unique(probe2gene)
# restrict to features annotated by a gene (since probe2gene was restricted to non NA values)
probe2proj <- probe2proj[names(probe2gene)]
# take median projection for genes annotated by several features
gene2proj <-
sapply(genesmanyfeatures,
function(gene, probe2proj, probe2gene) {
median(probe2proj[names(probe2gene[probe2gene %in% gene])])
}
, probe2proj = probe2proj, probe2gene = probe2gene
)
# merge with probe projection for genes annotated by only one probe
gene2proj2 <- probe2proj[!(names(probe2proj) %in% featuresmanygenes)]
names(gene2proj2) <- probe2gene[!(names(probe2gene) %in% featuresmanygenes)]
gene2proj <- c(gene2proj,gene2proj2)[orOrd]
gene2proj <- gene2proj[!is.na(gene2proj)]
return(gene2proj)
}
, probe2gene = probe2gene[!is.na(probe2gene)]
, genesmanyfeatures = genesmanyfeatures
, featuresmanygenes = featuresmanygenes
, gene2probe_manyfeatures = gene2probe_manyfeatures
)
gc()
dat_medianGenes <- lapply(gene2probe[genesmanyfeatures],
function (features, dat) {
# restriction to the probe sets annotating the genes
dataSub = dat[features,]
return(apply(dataSub, 2, median))
}
, dat = dat
)
dat_medianGenes = as.data.frame(t(as.data.frame(dat_medianGenes)))
rownames(dat_medianGenes) = as.character(genesmanyfeatures)
dat_genes = rbind(dat_genes, dat_medianGenes)
}
else {
genesComp <- llply(Slist(icaSet),
function(probe2proj, gene2probe) {
probe2proj <- probe2proj[gene2probe]
names(probe2proj) <- names(gene2probe)
return(probe2proj)
}
, gene2probe = gene2probe)
}
gc()
dat_genes <- dat_genes[names(genesComp[[1]]),]
icaSet@datByGene <- as.data.frame(dat_genes, stringsAsFactors = FALSE, check.names = FALSE)
icaSet@SByGene <- as.data.frame(genesComp, stringsAsFactors = FALSE, check.names = FALSE)
# redefine witness genes if they were selected based on a different annotation
if (length(witGenes(icaSet))>0 && length(intersect(witGenes(icaSet),geneNames(icaSet)))<nbComp(icaSet)) {
message(".....Redefine gene witnesses.....","\n")
witGenes(icaSet) <- selectWitnessGenes(icaSet=icaSet, params=params, level="gene", maxNbOcc=1, selectionByComp=NULL)
}
colnames(SByGene(icaSet)) <- colnames(S(icaSet))
gc()
return(icaSet)
}
##' This function annotates a set of features using \code{biomaRt}
##'
##' @title Annotation of features using \code{biomaRt}
##' @param features Feature IDs to be annotated
##' @param featureId The type of the feature IDs, in the \code{biomaRt} way (type \code{listFilters(mart)} to choose one)
##' @param geneId The type of the gene IDs, in the \code{biomaRt} way (type \code{listAttributes(mart)} to choose one)
##' @param mart The mart object (database and dataset) used for annotation, see function \code{useMart} of package \code{biomaRt}
##' @return A vector of gene IDs indexed by the feature IDs.
##' @author Anne Biton
##' @export
##' @examples if (interactive()) {
##' # define the database to be queried by biomaRt
##' mart <- useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
##'
##' # annotate a set of HG-U133a probe sets IDs into Gene Symbols
##' annotFeaturesWithBiomaRt(features = c("1007_s_at", "1053_at", "117_at", "121_at", "1255_g_at"),
##' featureId="affy_hg_u133a", geneId="hgnc_symbol", mart=mart)
##'
##' # annotate a set of Ensembl Gene IDs into Gene Symbols
##' annotFeaturesWithBiomaRt(features = c("ENSG00000101412", "ENSG00000112242",
##' "ENSG00000148773", "ENSG00000131747", "ENSG00000170312",
##' "ENSG00000117399"), featureId="ensembl_gene_id", geneId="hgnc_symbol", mart=mart)
##' }
annotFeaturesWithBiomaRt <- function(features,
featureId,
geneId,
mart = useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
) {
#input = features
featureId <- match.arg(featureId,c(NULL,listFilters(mart)[,1]))
#output = genes
geneId <- match.arg(geneId,c(NULL,listAttributes(mart)[,1]))
if (length(geneId)>1)
warning("Only the first element of 'geneID' will be used.")
if (length(featureId)>1)
warning("Only the first element of 'featureID' will be used.")
geneId <- geneId[1]
featureId <- featureId[1]
# probe -> gene
d <- getBM(attributes = c(featureId,geneId),
filters = featureId,
values = features,
mart = mart)
d <- subset(d, d[[geneId]] != "")
## features without available annotation which are not returned by getBM
noAnnot <- features[!(features %in% d[[featureId]])]
if (length(noAnnot)>0)
message("Number of features without associated annotation is ", length(noAnnot), "\n")
probe2gene <- as.character(d[[geneId]])
names(probe2gene) <- as.character(d[[featureId]])
probe2gene <- probe2gene[!is.na(probe2gene)]
allGenes <- unique(unlist(probe2gene))
return(probe2gene)
}
##' This function annotates IDs (typically gene IDs) provided by the user and returns an html file with their description.
##'
##'
##' \code{"hgnc_symbol", "ensembl_gene_id", "description", "chromosome_name", "start_position", "end_position", "band"}, and \code{"strand"},
##' are automatically added to the list of fields available in argument \code{typeRetrieved} queried on biomaRt.
##' The web-links to www.genecards.org and www.proteinatlas.org are automatically added in the columns of the output respectively
##' corresponding to \code{hgnc_symbol} and \code{ensembl_gene_id}.
##' @title Description of features using package \code{biomaRt}.
##' @param data Either a data.frame whose rownames or one of its columns contain the IDs to be annotated, or a vector of IDs.
##' @param filename The name of the HTML file where gene annotations are written.
##' @param mart Output of function \code{useMart} from package \code{biomaRt}.
##' @param typeId The type of IDs available in \code{data}, in the biomaRt way (type \code{listFilters(mart)} to choose one).
##' @param typeRetrieved The descriptors uses to annotate the features of \code{data} (type \code{listAttributes(mart)} to choose one or several).
##' @param sortBy Name of a column of \code{data} used to order the output.
##' @param sortAbs If TRUE absolute value of column \code{sortBy} is used to order the output.
##' @param colAnnot The column containing the IDs to be annotated, if NULL or missing and argument \code{data} is a data.frame, then rownames of \code{data} must contain the IDs.
##' @param decreasing If TRUE, the output is sorted by decreasing values of the \code{sortBy} column
##' @param highlight IDs to be displayed in colour red in the returned table
##' @param caption A title for the HTML table
##' @return This function returns a data.frame which contains annotations of the input data.
##' @examples
##' if (interactive()) {
##' ## define the database to be used
##' mart <- useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
##'
##' ### Describe:
##' ## a set of hgnc symbols with default descriptions (typeRetrieved=NULL)
##' genes <- c("TOP2A","E2F3","E2F1","CDK1","CDC20","MKI67")
##' writeGenes(data=genes, filename="foo", mart=mart, typeId = "hgnc_symbol")
##'
##' ## a data.frame indexed by hngc symbols, sort output according to column "values", add a title to the HTML output
##' datagenes <- data.frame(values=rnorm(6),row.names = genes)
##' writeGenes(data=datagenes, filename="foo", sortBy = "values", caption = "Description of some proliferation genes.")
##'
##' ## a set of Entrez Gene IDs with default descriptions
##' genes <- c("7153","1871","1869","983","991","4288")
##' writeGenes(data=genes, filename="foo", mart=mart, typeId = "entrezgene")
##' }
##' \dontrun{
##' ## add the GO category the genes belong to
##' ## search in listAttributes(mart)[,1] which filter correspond to the Gene Ontology -> "go_id"
##' writeGenes(data=genes, filename="foo", mart=mart, typeId = "entrezgene", typeRetrieved = "go_id")
##' }
##' @author Anne Biton
##' @export
##' @seealso \code{\link[biomaRt]{getBM}}, \code{\link[biomaRt]{listFilters}}, \code{\link[biomaRt]{listAttributes}}, \code{\link[biomaRt]{useMart}}
writeGenes <- function (
data
,filename = NULL
,mart = useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
,typeId = "hgnc_symbol"
,typeRetrieved=NULL #c("hgnc_symbol", "entrezgene")
,sortBy = NULL
,sortAbs = TRUE
,colAnnot = NULL
,decreasing = TRUE
,highlight = NULL
,caption = ""
) {
typeId <- match.arg(typeId,c(NULL,listFilters(mart)[,1]), several.ok = TRUE)
if (!is.null(typeRetrieved))
typeRetrieved <- match.arg(typeRetrieved,listAttributes(mart)[,1], several.ok = TRUE)
else
typeRetrieved <- c()
addDefaultDescr <- setdiff(c("description", "chromosome_name", "start_position", "end_position", "band","strand", "ensembl_gene_id"),c(typeRetrieved,typeId))
if (length(addDefaultDescr)>0)
typeRetrieved <- c(typeRetrieved,addDefaultDescr)
if (is.vector(data)) {
namesGenes <- data
data <- data.frame(row.names = namesGenes)
data[[typeId]] <- namesGenes
colData <- NULL
}
else {
namesGenes <- if (is.null(colAnnot)) rownames(data) else data[[colAnnot]]
data[[typeId]] <- namesGenes
colData <- colnames(data)
if (!is.null(sortBy))
colData <- colData[colData != sortBy]
}
## add symbol id by default
if (!("hgnc_symbol" %in% c(typeId, typeRetrieved))) {
typeRetrieved <- c("hgnc_symbol",typeRetrieved)
addSymbol <- TRUE
}
if (!("ensembl_gene_id" %in% c(typeId, typeRetrieved))) {
typeRetrieved <- c(typeRetrieved, "ensembl_gene_id")
addSymbol <- TRUE
}
d <- getBM(attributes = c(typeId, typeRetrieved),
filters = typeId,
values = namesGenes,
mart = mart)
if (!is.null(colData)) {
colData <- setdiff(colData,c(typeId,typeRetrieved))
if (length(colData)==0)
colData <- NULL
}
## add genes without available annotation which are not returned by getBM
noAnnot <- namesGenes[!(namesGenes %in% d[[typeId]])]
if (length(noAnnot) > 0) {
nr = nrow(d)
d[(nr+1):(nr+length(noAnnot)),typeId] <- noAnnot
}
## merge to original data
d <- merge(data, d, by = typeId, all.x = TRUE)
## Rank data frame according to the values in the sortBy column, or by start position
if (!is.null(sortBy)) {
if (!is.numeric(d[[sortBy]])) stop("The column used to sort the data.frame is not numeric")
if (sortAbs)
d <- d[order(abs(as.numeric(d[[sortBy]])), decreasing = decreasing),]
else
d <- d[order(as.numeric(d[[sortBy]]), decreasing = decreasing),]
d[[sortBy]] <- signif(d[[sortBy]],4)
d[[sortBy]] <- as.character(d[[sortBy]])
}
else d <- d[order(as.numeric(d[["start_position"]]),decreasing = FALSE),]
if (!is.null(filename)) {
dhtml = d
if (!is.null(highlight)) {
classGene <- rep("gene",times=length(dhtml$hgnc_symbol))
classGene[which(dhtml$hgnc_symbol %in% highlight)] <- "geneh"
dhtml$hgnc_symbol <- paste("<a class =", classGene, " href=\"http://www.genecards.org/cgi-bin/carddisp.pl?gene=", dhtml$hgnc_symbol, "\">", dhtml$hgnc_symbol,"</a>",sep = "")
}
else dhtml$hgnc_symbol <- paste("<a class = 'gene' href=\"http://www.genecards.org/cgi-bin/carddisp.pl?gene=",
dhtml$hgnc_symbol, "\">",
dhtml$hgnc_symbol,"</a>",
sep = "")
dhtml$ensembl_gene_id <- paste("<a class = 'normal' href='http://www.proteinatlas.org/gene_info.php?ensembl_gene_id=",
dhtml$ensembl_gene_id, "'>",
dhtml$ensembl_gene_id,"</a>",
sep = "")
dhtml = dhtml[,c(typeId, if(!is.null(colData)) c(sortBy, colData) else sortBy, if (!is.null(typeRetrieved)) typeRetrieved)]
x <- xtable(dhtml,
caption = caption,
align = c("l","l","l","c","c","l",rep("c",ncol(dhtml)-5)),
digits = 4)
x <- capture.output(print(x,
type = "html",
sanitize.text.function = force,
include.rownames=FALSE,
caption.placement = "top")
)
style <-
"
<head>
<style type='text/css'>
a.comp:link{text-decoration:none;color:white;}
a.comp:visited{text-decoration:none;color:white;}
a.comp:hover{text-decoration:none;font-weight:bold;color:black;}
a.gene:link{text-decoration:none;color:black;font-weight:bold;}
a.gene:visited{text-decoration:none;color:black;font-weight:bold;}
a.gene:hover{text-decoration:none;font-weight:bold;color:#B22222;}
a.geneh:link{text-decoration:none;color:red;font-weight:bold;}
a.geneh:visited{text-decoration:none;color:red;font-weight:bold;}
a.geneh:hover{text-decoration:none;font-weight:bold;color:red;}
a.normal:link{text-decoration:none;color:black;}
a.normal:visited{text-decoration:none;color:black;}
a.normal:hover{text-decoration:none;font-weight:bold;color:black;}
TH.geneTable {
padding: 2px;
border:2px solid;
color: white;
border-color: #8B1A1A;
text-align: center;
font-weight: bold;
background-color: #B22222
}
</style>
</head>"
x <- gsub(x,pattern="<TABLE",replacement = "<TABLE style='font-family:Helvetica; font-size:12px; border-top:solid thin black; border=2px;' ", ignore.case = TRUE)
x <- gsub(x,pattern="<CAPTION",replacement="<CAPTION style='color:black; text-align:left; font-size:14px;' ", ignore.case = TRUE)
x <- gsub(x,pattern="<TH>",replacement = "<TH class='geneTable'>", ignore.case = TRUE)
x <- paste(style,x,sep="")
x <- write(x, file = paste(gsub(filename,pattern=".htm",replacement="",fixed=TRUE),".htm",sep=""))
}
return(d)
}
##### =====================================
#### Write contributing genes for each
#### component in separate files
##### =====================================
writeProjByComp <- function(icaSet
### The IcaSet-object
,params
### The MineICAParams-object which contains the parameters of the analysis, the files are written in the path provided in the attribute \code{genesPath} of \code{params}. The attribute \code{selCutoff} is used to select the features/genes that are annotated and writte.
,mart = useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl")
### The mart object used for annotation, see function \code{useMart} of package biomaRt
,typeRetrieved=NULL #= listFilters(mart)[,1] #c("hgnc_symbol", "entrezgene")
### The type of the annotation you want to use to annotate the IDs of argument \code{data}
,addNbOcc = TRUE
### If TRUE, the number of components the features/genes contribute to is added to the output. A gene/feature is considered as a contributor of a component if its scaled projection value is higher than attribute \code{selCutoff} of \code{icaSet}.
,selectionByComp = NULL
### A list which contains the feature/gene projections on each component, already restricted to the ones considered as contributors.
,level = c("features","genes")
### The attribute of \code{icaSet} that will be annotated: either the feature projections, or the gene projections
,typeId
### The type of the IDs the features or the genes of \code{icaSet} correspond to. It must be provided in the biomaRt way (type listFilters(mart) to choose the typeId).
,selCutoffWrite=2.5
)
{
##details<< One file is created by component, each file is named by the index of the components (\code{indComp(icaSet)}) and located in the path \code{genePath(params)}.
## The genes are ranked according to their absolute projection values.
##
##
## In addition, this function write an html file named "genes2comp" which provides, for each feature/gene, the number of components in which the gene is contributor (according to the threshold \code{cutoffSel(params)}), and its projection value on all the components.
## The projection values are scaled.
##
sel <- i <- comp <- cutoff <- NULL
level <- match.arg(tolower(level), choices = c("genes","features"))
switch(level,
features={SlistByGene=Slist(icaSet); object="S"},
genes={SlistByGene=SlistByGene(icaSet); object="SByGene"}
)
if (missing(typeId))
typeId <- typeID(icaSet)[c(features="featureID_biomart",genes="geneID_biomart")[level]]
if (length(SlistByGene)==0)
stop(paste("The attribute",object,"is empty."))
system(paste("mkdir",genesPath(params)), ignore.stderr=TRUE)
selCutoff <- selCutoff(params)
if ((length(selCutoff)>1) & (length(selCutoff)<nbComp(icaSet))) {
warning("The length of 'selCutoff' must be either 1 or equal to the number of components. Only the first value is used.")
selCutoff <- selCutoff[1]
params["selCutoff"] <- selCutoff
}
if (is.null(selectionByComp))
selectionByComp <- selectContrib(SlistByGene, cutoff = selCutoff)
selectionByComp_write <- selectContrib(SlistByGene, cutoff = selCutoffWrite)
#### ******* Write one table by component with the gene projections and annotations
gene2comp = NULL
if (addNbOcc) {
## If selectionByComp is NULL, it will be recomputed by nbOccByGeneInComp
## sometimes we recompute the selection or we use a different selection from data because in the global script the done selection is made with a minimal cutoff in order to write a lot of genes
gene2comp <- nbOccByGeneInComp(sel = selectionByComp, cutoff = selCutoff)
gene2comp_nb <- unlist(lapply(gene2comp, length))
names(gene2comp_nb) <- names(gene2comp)
gene2comp_char <- as.character(lapply(gene2comp, paste, collapse = ",")); names(gene2comp_char) = names(gene2comp)
}
else {
gene2comp_nb <- NULL
gene2comp_char <- NULL
}
if (length(selCutoff)==1)
selCutoff <- rep(selCutoff,nbComp(icaSet))
annotD <-
foreach(sel=selectionByComp_write,i=1:length(SlistByGene)) %do% {
d <- data.frame(scaled_proj = sel);
rownames(d) <- names(sel);
if (addNbOcc) {
gene2comp_nb[is.na(gene2comp_nb)] <- 0
d[[paste("nbOcc_forThreshold:",selCutoff[i],sep="")]] <- "0"
d[[paste("comp_forThreshold:",selCutoff[i],sep="")]] <- ""
dbis <- d
d[intersect(rownames(d),names(gene2comp_nb)),paste("nbOcc_forThreshold:",selCutoff[i],sep="")] <-
paste("<a class='normal' href='genes2comp_threshold",
if (length(selCutoff)>1) paste(selCutoff,collapse="_") else selCutoff,".htm#",
intersect(rownames(d),names(gene2comp_nb)),"'>",
gene2comp_nb[intersect(rownames(d),names(gene2comp_nb))],"</a>",sep="")
dbis[intersect(rownames(d),names(gene2comp_nb)),paste("nbOcc_forThreshold:",selCutoff[i],sep="")] <- gene2comp_nb[intersect(rownames(d),names(gene2comp_nb))]
#d[[paste("nbOcc_forThreshold:",selCutoff,sep="")]][is.na(d[[paste("nbOcc_forThreshold:",selCutoff,sep="")]])] <- "0"
#dbis[[paste("nbOcc_forThreshold:",selCutoff,sep="")]][is.na(dbis[[paste("nbOcc_forThreshold:",selCutoff,sep="")]])] <- "0"
d[intersect(rownames(d),names(gene2comp_nb)),paste("comp_forThreshold:",selCutoff[i],sep="")] <-
sapply(gene2comp_char[intersect(rownames(d),names(gene2comp_nb))],
function(x) {
ssplit <- strsplit(x,split=",")[[1]]
#if (is.na(ssplit[1])) ssplit <- "0"
paste(paste("<a class='normal' href='",ssplit,".htm'>",ssplit,"</a>",sep=""),collapse = ",")
}
)
dbis[intersect(rownames(d),names(gene2comp_nb)),paste("comp_forThreshold:",selCutoff[i],sep="")] <- gene2comp_char[intersect(rownames(d),names(gene2comp_nb))]
}
if (length(sel) != 0) {
resW <- writeGenes(data = d,
filename = paste(genesPath(params),i,sep=""),
mart = mart,
typeId = typeId,
typeRetrieved = typeRetrieved,
sortBy = "scaled_proj",
caption = paste("<b>Component ",comp," : this table provides the scaled genes projections exceeding ",selCutoffWrite,". <br> The genes are ranked according to their scaled projection on the component.</b> <br>Click on the gene symbol and ensembl gene id to respectively access to the corresponding GeneCards and Protein Atlas pages. <br> Click on the components index to open the corresponding genes tables. Click on the number of occurrences of a gene to see its projections values on every component.", sep = ""))
if (addNbOcc) {
resW[[paste("comp_forThreshold:",selCutoff[i],sep="")]] <- dbis[[paste("comp_forThreshold:",selCutoff[i],sep="")]][match(resW[[typeId]], rownames(dbis))]
resW[[paste("nbOcc_forThreshold:",selCutoff[i],sep="")]] <- dbis[[paste("nbOcc_forThreshold:",selCutoff[i],sep="")]][match(resW[[typeId]], rownames(dbis))]
}
}
else
resW <- NULL
return(resW)
}
nbOcc.inallComp <- nbOccInComp (icaSet = icaSet,
params = params,
selectionByComp = NULL,
file = paste(genesPath(params),"/genes2comp_threshold",if (length(selCutoff)>1) paste(selCutoff,collapse="_") else selCutoff,".htm",sep = ""),
level = level)
nbOcc.inallComp <- as.data.frame(apply(apply(nbOcc.inallComp,2,gsub,pattern="<b>",replacement=""),2,gsub,pattern="</b>",replacement=""), stringsAsFactors=FALSE)
colnames(nbOcc.inallComp)[(ncol(nbOcc.inallComp)-nbComp(icaSet)+1):ncol(nbOcc.inallComp)] <- indComp(icaSet)
return(list(listAnnotComp=annotD, nbOccInComp=nbOcc.inallComp))
### This function returns a list of two elements: the first element is named 'listAnnotComp' and contain a list with the output of the function 'writeGenes' for each component, and the second element is named 'nbOccInComp' and provides a data.frame which gives for each feature/gene (row) its projection values across all the components (columns).
##seealso<<writeGenes, getBM, listFilters, listAttributes, useMart, selectContrib, nbOcc.inallComp
}
nbOccByGeneInComp <- function(
### For each feature/gene, this function returns the indices of the components they contribute to.
Slist,
### The list of components, each element contains the projection values
cutoff,
### The threshold used to define a gene as contributor
sel
### The list of components already restricted to the contributing genes
) {
if (missing(sel) || is.null(sel))
sel <- selectContrib(Slist, cutoff = cutoff)
names(sel) = NULL
## all genes (or features) that have a zvalue of projection higher than cutoffOnZval on components
allGenes = unique(unlist(lapply(sel,names)))
## for each gene, return the ids of the components it belongs to
genes2comp = lapply(allGenes,
function(gene, sel) {
isPresent = lapply(sel,
function(comp, gene) {
if (gene %in% names(comp)) return (TRUE)
else return(FALSE)
}
, gene = gene)
presentInComp = which(isPresent == TRUE)
return(presentInComp)
}
, sel = sel)
names(genes2comp) <- allGenes
return(genes2comp)
}
##' For each feature/gene, this function returns the indices of the components they contribute to.
##'
##' @title nbOccInComp_simple
##' @param icaSet An object of class \code{\link{IcaSet}}.
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the analysis. \code{cutoffSel(params)} is used as a threshold on the absolute projections to select the contributing features/genes.
##' @param selectionByComp The list of components already restricted to the contributing features/genes (each element is a vector of projection values indexed by features or genes).
##' @param level The attribute of \code{icaSet} to be used, the occurences of either the \code{"features"} (using \code{S(icaSet)}) or the \code{"genes"} (using \code{SByGene(icaSet)}) will be reported.
##' @return Returns a data.frame whose columns are: \code{gene} the feature or gene IDs, \code{nbOcc} the number of components the gene contributes to, \code{components} the indices of those components.
##' @author Anne Biton
##' @keywords internal
##' @examples
##' data(icaSetCarbayo)
##' params <- buildMineICAParams(resPath="carbayo/")
##' nbOcc <- MineICA:::nbOccInComp_simple(icaSet=icaSetCarbayo, params=params, level="genes")
##'
nbOccInComp_simple <- function(
icaSet,
params,
selectionByComp = NULL,
level = c("features","genes")
)
{
level <- match.arg(tolower(level), choices = c("features","genes"))
switch(level,
features={Slist=Slist(icaSet)},
genes={Slist=SlistByGene(icaSet)}
)
cutoff <- selCutoff(params)
if (is.null(selectionByComp))
selectionByComp <- selectContrib(Slist, cutoff = cutoff)
genesToComps_index <- nbOccByGeneInComp(Slist = Slist, sel = selectionByComp, cutoff = cutoff)
genesToComps_nb <- unlist(lapply(genesToComps_index, length));
names(genesToComps_nb) <- names(genesToComps_index)
genesToComps <- unlist(lapply(genesToComps_index, paste, collapse = ","));
names(genesToComps) <- names(genesToComps_index)
d.genesToComps <- data.frame(gene = names(genesToComps), nbOcc = genesToComps_nb[names(genesToComps)], components = as.character(genesToComps), stringsAsFactors = FALSE)
return(d.genesToComps)
### Returns a data.frame whose columns are: 'gene' the feature or gene ID, 'nbOcc' the number of components on which the gene contributes according to the threshold, 'components' the indices of these components.
}
##' selectWitnessGenes
##'
##' This function selects a gene per component.
##'
##' Selects as feature/gene witness, for each component, the first gene whose
##' absolute projection is greater than a given threshold in at the most
##' \code{maxNbOcc} components.
##' These witnesses can then be used as representatives of the expression behavior of the contributing genes of the components.
##'
##' When a feature/gene respecting the given constraints is not found, \code{maxNbOcc} is incremented of one until a gene is found.
##'
##'
##' @param icaSet An object of class \code{\link{IcaSet}}
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the analysis, the attribute \code{cutoffSel} is used as the threshold.
##' @param level The attribute of \code{icaSet} to be used, the witness
##' elements will be either selected within the \code{"features"} or the \code{"genes"}
##' @param maxNbOcc The maximum number of components where the genes can have an
##' absolute projection value higher than \code{cutoffSel(params)} in order to
##' be selected.
##' @param selectionByComp The list of components already restricted to the
##' contributing genes
##' @return This function returns a vector of IDs.
##' @export
##' @author Anne Biton
##' @examples
##' ## load an example of IcaSet
##' data(icaSetCarbayo)
##'
##' ## define parameters: features or genes are considered to be contributor
##' # when their absolute projection value exceeds a threshold of 4.
##' params <- buildMineICAParams(resPath="carbayo/", selCutoff=4)
##'
##' ## selection, as gene witnesses, of the genes whose absolute projection is greater than 4
##' # in at the most one component. I.e, a gene is selected as a gene witness of a component
##' # if he has a large projection on this component only.
##' selectWitnessGenes(icaSet=icaSetCarbayo, params=params, level="genes", maxNbOcc=1)
##'
##' ## selection, as gene witnesses, of the genes whose absolute projection is greater than 4
##' # in at the most two components.
##' # I.e, a gene is selected as a gene witness of a given component if he has a large projection
##' # in this component and at the most another.
##' selectWitnessGenes(icaSet=icaSetCarbayo, params=params, level="genes", maxNbOcc=2)
##'
##'
selectWitnessGenes <- function(icaSet,
params,
level = c("genes","features"),
maxNbOcc = 1,
selectionByComp = NULL
) {
level <- match.arg(level)
if ((maxNbOcc) < 1)
stop("maxNbOcc must at least equal 1.")
switch(level,
features={Slist=Slist(icaSet)},
genes={Slist=SlistByGene(icaSet)}
)
cutoff <- selCutoff(params)
if (is.null(selectionByComp) | missing(selectionByComp))
selectionByComp <- selectContrib(Slist, cutoff = cutoff)
## For components with no genes selected using the cutoff,
indEmptySel <- which(unlist(lapply(selectionByComp,length))==0)
selectionByComp[indEmptySel] <- Slist[indEmptySel]
docc <- nbOccInComp_simple(icaSet = icaSet, params = params, selectionByComp = selectionByComp)
nbocc <- docc$nbOcc
names(nbocc) <- docc[,1]
## Select as witness genes the first gene whose contribution is greater than cutoff in at the most maxNbOcc components.
witGenes <-
lapply(selectionByComp,
function(sel, nbocc, maxNbOcc) {
sel <- sort(abs(sel), decreasing = TRUE)
ind <- which(nbocc[names(sel)] <= maxNbOcc)
while (is.na(ind[1])) {
maxNbOcc <- maxNbOcc+1
ind <- which(nbocc[names(sel)] <= maxNbOcc)
}
return(names(sel[ind[1]]))
}, nbocc = nbocc
, maxNbOcc = maxNbOcc
)
return(unlist(witGenes))
### This function returns a vector of IDs.
}
##' For each feature/gene, this function returns the components they contribute
##' to and their projection values across all the components.
##'
##' A feature/gene is considered as a contributor when its scaled projection value exceeds the threshold \code{selCutoff(icaSet)}.
##'
##' This function plots the number of times the feature/gene is a contributor as a function of the standard deviation of its expression profile.
##'
##' The created files are located in \code{genePath(params)}. An extensiom '.htm' and '.pdf' is respectively added to the \code{file} name for the data.frame and the plot outputs.
##' @title Select components the features contribute to
##' @param icaSet An object of class \code{\link{IcaSet}}
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the
##' analysis, the attribute \code{cutoffSel} is used as a threshold on the
##' absolute projections to determine which genes contribute to the components.
##' @param selectionByComp The list of components already restricted to the
##' contributing genes
##' @param level The attribute of \code{icaSet} to be used, are reported the
##' occurences of either the \code{"features"} or the \code{"genes"}.
##' @param file The file where the output data.frame and plots are written.
##' @return Returns a data.frame whose columns are: 'gene' the feature or gene
##' ID, 'nbOcc' the number of components on which the gene contributes according
##' to the threshold, 'components' the indices of these components, and then the
##' component indices which contain its projection values.
##' @export
##' @author Anne Biton
##' @examples
##' data(icaSetCarbayo)
##' params <- buildMineICAParams(resPath="carbayo/")
##' nbOcc <- nbOccInComp(icaSet=icaSetCarbayo, params=params, level="genes", file="gene2MixingMatrix")
##'
nbOccInComp <- function(
icaSet
,params
,selectionByComp = NULL
,level = c("features","genes")
,file = NULL
)
{
level <- match.arg(tolower(level), choices = c("features","genes"))
switch(level,
features={Slist=Slist(icaSet)},
genes={Slist=SlistByGene(icaSet)}
)
##Slist <- icaSet[object]
cutoff <- selCutoff(params)
if (is.null(selectionByComp))
selectionByComp <- selectContrib(Slist, cutoff = cutoff)
d.genesToComps <- nbOccInComp_simple(icaSet = icaSet, params = params, selectionByComp = selectionByComp, level = level)
if (is.null(datByGene(icaSet)) | nrow(datByGene(icaSet))==0)
dat <- read.table (datfile(params), header = TRUE)
else {
if (level == "genes")
dat <- datByGene(icaSet)
else
dat <- assayDataElement(icaSet,"dat")
}
dat.sd = getSdExpr(d.genesToComps$gene, dat)
d.genesToComps$sd_expr = signif(dat.sd[d.genesToComps$gene],5)
d.genesToComps = d.genesToComps[order(d.genesToComps$sd_expr, decreasing = TRUE),]
## plot sd expr vs projection
d.aggregate <- aggregate(d.genesToComps[,c("nbOcc","sd_expr")],by=list(occ=as.factor(d.genesToComps$nbOcc)),FUN=mean)
pdf(file=gsub(file,pattern="htm",replacement="pdf"),width = 6, height = 6, title=gsub(file,pattern="htm",replacement="pdf"))
if (anyDuplicated(d.aggregate$nbOcc) != 0)
d.aggregate <- d.aggregate[-which(duplicated(d.aggregate$nbOcc)),]
plot(d.aggregate$sd_expr,
d.aggregate$nbOcc,
pch = 19,
col = "black",
ylab = "number of occurence",
xlab = "mean of sd(expr)" ,
type = "b",
main = paste("Number of gene contributions to a component \n as a function of the average standard deviation of their profiles","\n cutoff ",if (length(unique(cutoff))>1) paste(cutoff,collapse=",") else cutoff, sep=""),
cex.main = 1)
dev.off()
if (!is.null(file)) {
## merge with a table providing the genes scaled projections on the components
compsel <- proj <- NULL
genesProj <- foreach(proj = Slist, compsel = selectionByComp, .combine = cbind) %dopar% {
# proj of the contributing genes
p <- signif(proj[d.genesToComps$gene],4)
names(p) <- d.genesToComps$gene
# mark the contributing genes in an other color(i.e. genes exceeding the threshold)
p[names(compsel)] <- paste("<b>",p[names(compsel)],"</b>", sep = "")
return(p)
}
genesProj <- as.data.frame(genesProj, row.names = d.genesToComps$gene)
d.genesToCompsbis <-cbind(d.genesToComps, genesProj[d.genesToComps$gene,])
names(genesProj) <- paste("<a class='comp' href='",c(1:ncol(genesProj)),".htm'>",c(1:length(Slist)),"</a>",sep="")
d.genesToComps <- cbind(d.genesToComps, genesProj[d.genesToComps$gene,])
d.genesToComps$gene <- paste("<a class='gene' name='",d.genesToComps$gene,"' href='http://www.genecards.org/cgi-bin/carddisp.pl?gene=", d.genesToComps$gene, "'>", d.genesToComps$gene,"</a>",sep = "")
d.genesToComps$components <-
sapply(as.character(d.genesToComps$components),
function(x) {
ssplit <- strsplit(x,split=",")[[1]]
paste(paste("<a class='normal' href='",ssplit,".htm'>",ssplit,"</a>",sep=""),collapse = ",")
})
style <-
"
<head>
<style type='text/css'>
a.comp:link{text-decoration:none;color:white;}
a.comp:visited{text-decoration:none;color:white;}
a.comp:hover{text-decoration:none;font-weight:bold;color:black;}
a.gene:link{text-decoration:none;color:black;font-weight:bold;}
a.gene:visited{text-decoration:none;color:black;font-weight:bold;}
a.gene:hover{text-decoration:none;font-weight:bold;color:#B22222;}
a.normal:link{text-decoration:none;color:black;}
a.normal:visited{text-decoration:none;color:black;}
a.normal:hover{text-decoration:none;font-weight:bold;color:black;}
TH.geneTable {
padding: 2px;
border:2px solid;
color: white;
border-color: #8B1A1A;
text-align: center;
font-weight: bold;
background-color: #B22222
}
</style>
</head>"
x <- xtable(d.genesToComps, caption = paste("<b>This table describes the genes contributing to at least one component, given the threshold(s) ", if (length(cutoff>1)) paste(cutoff,collapse=",") else cutoff, ". <br> The genes are ranked according to the standard deviation of their expression profiles.</b> <br>Click on the gene id to access to its GeneCards page. <br> Click on the components index to access the scaled projections of the complete list of genes. ", sep = ""),align = c("l","l","c","l",rep("l",ncol(d.genesToComps)-3)))
x <- capture.output(print( x,
type = "html",
sanitize.text.function = force,
include.rownames=FALSE,
caption.placement = "top")
)
x <- gsub(x,pattern="<TABLE",replacement = "<TABLE style='font-family:Helvetica; font-size:12px; border-top:solid thin black; border=2px;' ", ignore.case = TRUE)
x <- gsub(x,pattern="<CAPTION",replacement="<CAPTION style='color:black; text-align:left; font-size:14px;' ", ignore.case = TRUE)
x <- gsub(x,pattern="<TH>",replacement = "<TH class='geneTable'>", ignore.case = TRUE)
x <- paste(style,x,sep="")
write(x, file = file)
}
else {
d.genesToComps <- cbind(d.genesToComps, as.data.frame(Slist)[d.genesToComps$gene,])
colnames(d.genesToComps)[ncol(d.genesToComps):(ncol(d.genesToComps)-length(compNames(icaSet)))] <- rev(compNames(icaSet))
}
return(d.genesToCompsbis)
}
##' getSdExpr
##'
##' Compute standard deviation of the gene expression
##'
##'
##' @param features IDs
##' @param dat Expression data indexed by IDs
##' @return Returns a vector
##' @author Anne Biton
##' @export
##' @keywords internal
##' @examples
##' dat <- matrix(rnorm(1000),ncol=10,nrow=100)
##' rownames(dat) <- 1:100
##' MineICA:::getSdExpr(features = 2:20, dat = dat)
##'
getSdExpr <- function (
### Compute standard deviation of the gene expression
features
### IDs
,dat
### Expression data indexed by IDs
) {
genes.sdExpr <- as.numeric( sapply(features, function (gene, dat) { return(sd(as.numeric(dat[gene,]))) }, dat))
names(genes.sdExpr) <- features
return(genes.sdExpr)
### Returns a vector
}
##' This function builds an object of class \code{\link[MineICA:MineICAParams-class]{MineICAParams}}.
##' It contains the parameters that will be used by function \code{\link{runAn}} to analyze the ICA decomposition contained in an object of class \code{\link[MineICA:IcaSet-class]{IcaSet}}.
##'
##'
##' @title Creates an object of class MineICAParams
##' @param Sfile A txt file containing the Source matrix S.
##' @param Afile A txt file containing the Mixing matrix A.
##' @param datfile A txt file containing the data (e.g expression data) on which the decomposition was calculated.
##' @param annotfile Either a "rda" or "txt" file containing the annotation data for the samples (must be of dimensions samples x annotations).
##' @param resPath The path where the outputs of the analysis will be written, default is the current directory.
##' @param genesPath The path _within_ the resPath where the gene projections will be written. If missing, will be automatically attributed as \code{resPath}/ProjByComp/.
##' @param annot2col A vector of colors indexed by annotation levels. If missing, will be automatically attributed using function \code{annot2Color}.
##' @param pvalCutoff The cutoff used to consider a p-value significant, default is 0.05.
##' @param selCutoff The cutoff applied to the absolute feature/gene projection values to consider them as contributors, default is 3. Must be either of length 1 and the same treshold is applied to all components, or of length equal to the number of components in order to a specific threshold is for each component.
##' @return An object of class \code{\link{MineICAParams}}
##' @export
##' @author Anne Biton
##' @seealso \code{\link[MineICA:MineICAParams-class]{MineICAParams}}, \code{\link[MineICA]{runAn}}
##' @examples
##'
##' ## define default parameters and fill resPath
##' params <- buildMineICAParams(resPath="resMineICACarbayo/")
##'
##' ## change the default cutoff for selection of contribugint genes/features
##' params <- buildMineICAParams(resPath="resMineICACarbayo/", selCutoff=4)
##'
buildMineICAParams <- function (Sfile = new("character"), Afile=new("character"), datfile= new("character"), annotfile=new("character"), resPath="", genesPath, annot2col=new("character"), pvalCutoff= 0.05, selCutoff=3) {
if (resPath != "") {
home <- system("echo $HOME",intern=TRUE);
resPath <- gsub(resPath,pattern="~",replacement=home);
}
if (missing(genesPath))
genesPath <- paste(resPath,"ProjByComp/",sep="")
else
genesPath <- paste(resPath,genesPath,sep="/")
params <- new("MineICAParams",
Sfile = Sfile,
Afile= Afile,
datfile= datfile,
annotfile=annotfile,
resPath= resPath ,
genesPath=genesPath,
annot2col=annot2col,
pvalCutoff=pvalCutoff,
selCutoff=selCutoff)
if (resPath != "")
system(paste("mkdir",resPath(params)), ignore.stderr=TRUE)
system(paste("mkdir",genesPath(params)), ignore.stderr=TRUE)
return(params)
}
##' This function builds an object of class \code{\link{IcaSet}}.
##'
##'
##' @param params An object of class \code{\link{MineICAParams}} containing the parameters of the analysis
##' @param A The mixing matrix of the ICA decomposition (of dimension samples x components).
##' @param S The source matrix of the ICA decomposition (of dimension features x components).
##' @param dat The data matrix the ICA was applied to (of dimension features x samples).
##' @param pData Phenotype data, a data.frame which contains the sample
##' informations of dimension samples x annotations.
##' @param fData Feature data, a data.frame which contrains the feature
##' descriptions of dimensions features x annotations.
##' @param witGenes A vector of witness genes. They are representative of the
##' expression behavior of the contributing genes of each component.
##' If missing or NULL, they will be automatically attributed using function \code{\link{selectWitnessGenes}}.
##' @param compNames A vector of component labels.
##' @param refSamples A vector of reference sample IDs (e.g the "normal" samples).
##' @param annotation An annotation package (e.g a ".db" package specific to the
##' microarray used to generate \code{dat})
##' @param chipManu If microarray data, the manufacturer: either 'affymetrix' or 'illumina'.
##' @param chipVersion For illumina microarrays: the version of the microarray.
##' @param alreadyAnnot TRUE if the feature IDs contained in the row names of \code{dat} and \code{S} already correspond to the final level of annotation (e.g if they are already gene IDs). In that case, no annotation is performed.
##' @param typeID A character vector specifying the annotation IDs, it includes
##' three elements : \describe{
##' \item{geneID_annotation}{the IDs from the
##' package to be used to annotate the features into genes. It will be used to
##' fill the attributes \code{datByGene} and \code{SByGene} of the \code{icaSet}.
##' It must match one of the objects the corresponding package supports
##' (you can access the list of objects by typing ls("package:packagename")). If
##' no annotation package is provided, this element is not useful.}
##' \item{geneID_biomart}{the type of gene IDs, as available in
##' \code{listFilters(mart)}; where mart is specified as described in \code{\link[biomaRt]{useMart}}.
##' If you have directly built the IcaSet at the
##' gene level (i.e if no annotation package is used), \code{featureID_biomart} and
##' \code{geneID_biomart} will be identical.}
##' \item{featureID_biomart}{the
##' type of feature IDs, as available in \code{listFilters(mart)}; where
##' \code{mart} is specified as described in function \code{\link[biomaRt]{useMart}}.
##' Not useful if you work at the gene level.} }
##' @param runAnnot If TRUE, \code{icaSet} is annotated with function \code{annotInGene}.
##' @param organism The organism the data correspond to.
##' @param mart The mart object (database and dataset) used for annotation, see function \code{useMart} of package \code{biomaRt}
##' @seealso \code{\link{selectWitnessGenes}}, \code{\link{annotInGene}}
##' @return An object of class IcaSet
##' @export
##' @author Anne Biton
##' @examples
##'
##' dat <- data.frame(matrix(rnorm(10000),ncol=10,nrow=1000))
##' rownames(dat) <- paste("g", 1:1000, sep="")
##' colnames(dat) <- paste("s", 1:10, sep="")
##'
##' ## build a data.frame containing sample annotations
##' annot <- data.frame(type=c(rep("a",5),rep("b",5)))
##' rownames(annot) <- colnames(dat)
##'
##' ## run ICA
##' resJade <- runICA(X=dat, nbComp=3, method = "JADE")
##'
##' ## build params
##' params <- buildMineICAParams(resPath="toy/")
##'
##' ## build IcaSet object
##' icaSettoy <- buildIcaSet(params=params, A=data.frame(resJade$A), S=data.frame(resJade$S),
##' dat=dat, pData=annot, alreadyAnnot=TRUE)
##' params <- icaSettoy$params
##' icaSettoy <- icaSettoy$icaSet
##'
##' \dontrun{
##' ## load data
##' library(breastCancerMAINZ)
##' data(mainz)
##'
##' ## run ICA
##' resJade <- runICA(X=dataMainz, nbComp=10, method = "JADE", maxit=10000)
##'
##' ## build params
##' params <- buildMineICAParams(resPath="mainz/")
##'
##' ## build IcaSet object
##'
##' # fill typeID, Mainz data originate from affymetrix HG-U133a microarray and are indexed by probe sets
##' # we want to annotate the probe sets into Gene Symbols
##' typeIDmainz <- c(geneID_annotation="SYMBOL", geneID_biomart="hgnc_symbol", featureID_biomart="affy_hg_u133a")
##'
##' icaSetMainz <- buildIcaSet(params=params, A=data.frame(resJade$A), S=data.frame(resJade$S),
##' dat=exprs(mainz), pData=pData(mainz),
##' annotation="hgu133a.db", typeID= c(geneID_annotation = "SYMBOL",
##' geneID_biomart = "hgnc_symbol", featureID_biomart = "affy_hg_u133a"),
##' chipManu = "affymetrix", runAnnot=TRUE,
##' mart=useMart(biomart="ensembl", dataset="hsapiens_gene_ensembl"))
##' }
buildIcaSet <- function(params,
A,
S,
dat,
pData = new("data.frame"),
fData = new("data.frame"),
witGenes= new("character"),
compNames= new("character"),
refSamples= new("character"),
annotation = new("character"),
chipManu = new("character"),
chipVersion = new("character"),
alreadyAnnot = FALSE,
typeID = c(geneID_annotation = "SYMBOL", geneID_biomart = "hgnc_symbol", featureID_biomart = ""),
runAnnot = TRUE,
organism = "Human",
mart=new("Mart")
) {
if (missing(dat)) {
if (is.null(datfile(params)) | datfile(params) == "" | length(datfile(params))==0)
stop("Either the expression matrix or the corresponding file must be provided.")
dat <- read.table(datfile(params), header = TRUE, stringsAsFactors= FALSE, check.names = FALSE)
}
if (missing(pData) & (length(annotfile(params))>0)) {
if (length(grep(pattern = ".rda", x = tolower(annotfile(params))))>0)
pData <- get(load(annotfile(params)))
else
pData <- read.table(annotfile(params), header = TRUE, stringsAsFactors= FALSE, check.names = FALSE)
}
if (missing(A))
A <- readA(Afile = Afile(params), dat = dat) #, sep = sep)
if (missing(S))
S <- readS(Sfile = Sfile(params), dat = dat) #, sep = sep)
if(length(compNames) <= ncol(A))
compNames <- as.character(1:ncol(A))
if (length(annotation)==0 || (annotation == "")) {
if (is.null(mart))
stop("A 'mart' object must be defined using function 'useMart'.")
}
icaSet <- new("IcaSet",
A = A,
S = S,
##witGenes = witGenes, ## to be initialized after annotation in gene
annotation=annotation,
dat=as.matrix(dat),
compNames=compNames,
chipManu = chipManu,
chipVersion=chipVersion,
refSamples=refSamples,
indComp = 1:ncol(A),
typeID = typeID,
mart=mart)
if (!missing(pData) || ncol(pData)>0)
pData(icaSet) <- pData
organism(icaSet) <- organism
annot2col(params) <- annot2Color(pData)
if (alreadyAnnot) {
SByGene(icaSet) <- S
datByGene(icaSet) <- dat
runAnnot <- FALSE
}
else {
SByGene(icaSet) <- datByGene(icaSet) <- data.frame()
}
if (runAnnot) {
if (length(annotation(icaSet))==0 || (annotation(icaSet) == "")) {
if (missing(mart) || is.null(mart))
stop("A 'mart' object must be defined using function 'useMart'.")
#warning("Since no annotation package is provided, annotation will be performed using biomaRt.")
#stop(paste("You must provide an 'annotation' (i.e package) attribute for the annotation."))
}
else {
if (!("geneID_annotation" %in% names(typeID(icaSet))))
stop(paste("Since you want to annotate the features using ", annotation(icaSet),", typeID attribute of IcaSet object must contain an element 'geneID_annotation' which determines the object from the package you want to use."))
if ((length(chipManu(icaSet))>1 && chipManu(icaSet) == "illumina") & (!(toupper(typeID(icaSet)["geneID_annotation"]) %in% c("ENTREZID", "SYMBOL") )))
stop("For illumina, features can only be annotated in ENTREZID or SYMBOL. You will need to annotate yourself the IcaSet object if you want to use different IDs.")
}
}
## if no annotation package is provided, still the biomart ID of the features or genes must be provided
if (length(annotation(icaSet))==0 || (annotation(icaSet) == "")) {
if (length(typeID(icaSet)["geneID_biomart"]) == 0 || typeID(icaSet)["geneID_biomart"] == "" || is.na(typeID(icaSet)["geneID_biomart"]))
if (length(typeID(icaSet)["featureID_biomart"]) == 0 || typeID(icaSet)["featureID_biomart"] == "" || is.na(typeID(icaSet)["featureID_biomart"]) )
stop("Since no annotation package is available, you must fill at least 'featureID_biomart' or 'geneID_biomart' in typeID attribute.")
else
typeID(icaSet)["geneID_biomart"] <- typeID(icaSet)["featureID_biomart"]
else
if (length(typeID(icaSet)["featureID_biomart"]) == 0 || typeID(icaSet)["featureID_biomart"] == "" || is.na(typeID(icaSet)["featureID_biomart"]))
typeID(icaSet)["featureID_biomart"] <- typeID(icaSet)["geneID_biomart"]
}
if (length(chipManu(icaSet))>1 && chipManu(icaSet) == "illumina" & runAnnot) {
message("Since you are using illumina microarrays, additional packages are needed to complete the annotation: lumi, lumiHumanAll.db and either lumiRatIDMapping or lumiMouseIDMapping depending on the organism. \n")
message("...You must load packages 'lumi' and 'lumiHumanAll.db if not done yet'...")
}
## launched even if runAnnot = FALSE in order to write rnk files if needed
icaSet <- annotInGene(icaSet = icaSet, params = params, annot = runAnnot)
witGenes(icaSet) <- witGenes
if (length(witGenes(icaSet))==0)
witGenes(icaSet) <- selectWitnessGenes(icaSet=icaSet, params=params, level = if (nrow(SByGene(icaSet))>0) "genes" else "features", maxNbOcc = 1, selectionByComp = NULL)
return(list(icaSet=icaSet,params = params))
}
##' This function selects the features having the largest Inter Quartile Range (IQR).
##'
##' @title Selection of features based on their IQR
##' @param data Measured data of dimension features x samples (e.g, gene expression data)
##' @param nb The number of features to be selected
##' @return A subset of \code{data} restricted to the features having the \code{nb} highest IQR value
##' @author Pierre Gestraud
##' @export
##' @examples
##' dat <- matrix(rnorm(10000),ncol=10,nrow=1000)
##' rownames(dat) <- 1:1000
##' selectFeatures_IQR(data=dat, nb=500)
##'
selectFeatures_IQR <-
function(data, nb){
iqrValues <- apply(data, 1, IQR, na.rm=TRUE)
sortedIqr <- sort(iqrValues, decreasing=TRUE)
maxIqr <- sortedIqr[nb]
index <- which(iqrValues >= maxIqr)
nbGenes <- length(index)
message("Number of selected genes is ", nbGenes, "\n")
message("Max IQR is ", signif(maxIqr,4), "\n")
return(data[index,])
}
##' Writes the gene projection values of each component in a '.rnk' file for GSEA.
##'
##' The .rnk format requires two columns, the first containing the gene IDs, the second containing the projection values.
##' The genes are ordered by projection values.
##' The files are named "index-of-component_abs.rnk" if \code{abs=TRUE}, or "index-of-component.rnk" if \code{abs=FALSE}.
##' @title Write rnk files containing gene projections
##' @param icaSet An object of class \code{IcaSet}
##' @param abs If TRUE (default) the absolute projection values are used.
##' @param path The path that will contain the rnk files.
##' @return NULL
##' @author Anne
##' @export
writeRnkFiles <- function(icaSet, abs=TRUE, path) {
dirGsea <- paste(path,"/GSEA/",sep="")
system(paste("mkdir",dirGsea), ignore.stderr=TRUE)
Slist <- SlistByGene(icaSet)
if (abs) {
system(paste("mkdir ",dirGsea, "rnk_abs",sep=""), ignore.stderr=TRUE)
rnk_path= paste(dirGsea,"rnk_abs/",sep="")
}
else {
system(paste("mkdir ",dirGsea, "rnk",sep=""), ignore.stderr=TRUE)
rnk_path= paste(dirGsea,"rnk/",sep="")
}
if (abs)
Slist <- lapply(Slist, function(x) {abs(x)})
sapply (1:length(Slist),
function (indcomp, path, Slist, icaSet) {
comp.hugo <- Slist[[indcomp]]
dat = as.data.frame(comp.hugo)
write.table(dat,file = paste(path,"/",indComp(icaSet)[indcomp],"_",if(abs) "abs",".rnk",sep=""), sep = "\t", quote = FALSE, col.names = FALSE)
}
, path = rnk_path
, Slist = Slist
, icaSet = icaSet
)
}
