Nothing
#' Read the raw mutation data of Mutation Position Format.
#'
#' @description
#' The mutation position format is tab-delimited text file, where
#' the 1st-5th columns shows sample names, chromosome names,
#' coordinates, reference bases (A, C, G, or T) and
#' the alternate bases (A, C, G, or T), respectively. An example is as follows;
#'
#' ---
#'
#' sample1 chr1 100 A C
#'
#' sample1 chr1 200 A T
#'
#' sample1 chr2 100 G T
#'
#' sample2 chr1 300 T C
#'
#' sample3 chr3 400 T C
#'
#' ---
#'
#' Also, this function usually can accept compressed files (e.g., by gzip, bzip2
#' and so on) when using recent version of R.
#'
#' @param infile the path for the input file for the mutation data of Mutation
#' Position Format.
#' @param numBases the number of upstream and downstream flanking bases
#' (including the mutated base) to take into account.
#' @param trDir the index representing whether transcription direction is
#' considered or not.
#' The gene annotation information is given by UCSC knownGene
#' (TxDb.Hsapiens.UCSC.hg19.knownGene object) When trDir is TRUE, the mutations
#' located in intergenic region are excluded from the analysis.
#' @param bs_genome this argument specifies the reference genome (e.g., B
#' Sgenome.Mmusculus.UCSC.mm10 can be used for the mouse genome).
#' See https://bioconductor.org/packages/release/bioc/html/BSgenome.html for the
#' available genome list
#' @param txdb_transcript this argument specified the transcript database
#' (e.g., TxDb.Mmusculus.UCSC.mm10.knownGene can be used for the mouse genome).
#' See https://bioconductor.org/packages/release/bioc/html/AnnotationDbi.html
#' for details.
#' @return The output is an instance of MutationFeatureData S4 class (which
#' stores summarized information on mutation data). This will be typically used
#' as the initial values for the global test and the local test.
#'
#' @importFrom utils read.table
#' @importFrom methods new
#' @examples
#' inputFile <- system.file("extdata/esophageal.mp.txt.gz", package="HiLDA")
#' G <- hildaReadMPFile(inputFile, numBases=5, trDir=TRUE)
#'
#'
#' @export
hildaReadMPFile <- function(infile, numBases = 3, trDir = FALSE,
bs_genome = NULL, txdb_transcript = NULL) {
if (is.null(bs_genome) == TRUE) {
bs_genome <- BSgenome.Hsapiens.UCSC.hg19::BSgenome.Hsapiens.UCSC.hg19
}
if (is.null(txdb_transcript) == TRUE) {
txdb_transcript <-
TxDb.Hsapiens.UCSC.hg19.knownGene::TxDb.Hsapiens.UCSC.hg19.knownGene
}
fdim <- c(6, rep(4, numBases - 1), rep(2, as.integer(trDir)))
type <- "independent"
if (numBases %% 2 != 1) {
stop("numBases should be odd numbers")
}
centerInd <- (numBases + 1) / 2
mutFile <- read.table(infile, sep="\t", header=FALSE,
stringsAsFactors = FALSE)
chrInfo <- mutFile[,2]
posInfo <- mutFile[,3]
ref_base <- Biostrings::DNAStringSet(mutFile[,4])
alt_base <- Biostrings::DNAStringSet(mutFile[,5])
sampleName_str <- as.character(mutFile[,1])
gr <- GenomicRanges::makeGRangesFromDataFrame(data.frame(chr = chrInfo,
start = posInfo,
end = posInfo),
ignore.strand = TRUE)
ranges <- GenomicRanges::resize(gr, numBases, fix = "center")
context <- Biostrings::getSeq(bs_genome, ranges)
## check the consistency between the input reference base and the obtained
## base using hg19 reference genome.
removeInd <- which(XVector::subseq(context, start = centerInd,
end = centerInd) != ref_base)
if (length(removeInd) > 0) {
warning("The central bases are inconsistent in ", length(removeInd),
" mutations. We have removed them.")
for(varname in c("context", "ref_base", "alt_base",
"sampleName_str", "chrInfo", "posInfo")) {
assign(varname, get(varname)[-removeInd])
}
}
# check the characters on alternative base
alphabetFreq <- Biostrings::alphabetFrequency(alt_base)
removeInd <- which(rowSums(alphabetFreq[,seq_len(4)]) != 1)
if (length(removeInd) > 0) {
warning("The characters other than (A, C, G, T) are included in
alternate bases of ", length(removeInd),
" mutations. We have removed them.")
for(varname in c("context", "ref_base", "alt_base",
"sampleName_str", "chrInfo", "posInfo")) {
assign(varname, get(varname)[-removeInd])
}
}
# check the characters on flanking bases
alphabetFreq <- Biostrings::alphabetFrequency(context)
removeInd <- which(alphabetFreq[,"A"] + alphabetFreq[,"C"] +
alphabetFreq[,"G"] + alphabetFreq[,"T"] != numBases)
if (length(removeInd) > 0) {
lapply(list(context, ref_base, alt_base,
sampleName_str, chrInfo, posInfo),
function(obj) obj <- obj[-removeInd])
warning("The characters other than (A, C, G, T) are included in
flanking bases of ", length(removeInd),
" mutations. We have removed them.")
}
# check the characters on alternative base
alphabetFreq <- Biostrings::alphabetFrequency(alt_base)
removeInd <- which(ref_base == alt_base)
if (length(removeInd) > 0) {
warning("The reference base and alternative bases are equal for ",
length(removeInd), " mutations. We have removed them.")
for(varname in c("context", "ref_base", "alt_base",
"sampleName_str", "chrInfo", "posInfo")) {
assign(varname, get(varname)[-removeInd])
}
}
revCompInd <- which(as.character(XVector::subseq(context,
start = centerInd,
end = centerInd)) %in%
c("A", "G"))
context[revCompInd] <- Biostrings::reverseComplement(context[revCompInd])
ref_base[revCompInd] <- Biostrings::reverseComplement(ref_base[revCompInd])
alt_base[revCompInd] <- Biostrings::reverseComplement(alt_base[revCompInd])
# Obtaining transcription strand information using GenomicRanges packages
if (trDir == TRUE) {
gr <- GenomicRanges::makeGRangesFromDataFrame(data.frame(chr = chrInfo,
start = posInfo,
end = posInfo),
ignore.strand = TRUE)
txdb <- txdb_transcript
txdb_bed <- GenomicFeatures::transcripts(txdb)
gr_txdb <- GenomicRanges::findOverlaps(gr, txdb_bed,
ignore.strand = FALSE)
gr_strand <- cbind(
S4Vectors::queryHits(gr_txdb),
as.character(S4Vectors::as.factor(
BiocGenerics::strand(txdb_bed[S4Vectors::subjectHits(gr_txdb)]))))
ugr_strand <- unique(gr_strand[gr_strand[, 2] != "*" ,], MARGIN=1)
rmdup_ugr_strand <- ugr_strand[!duplicated(ugr_strand[, 1]), ]
txdb_plus_gr_ind <- as.integer(
rmdup_ugr_strand[rmdup_ugr_strand[, 2] == "+", 1])
txdb_minus_gr_ind <-
as.integer(rmdup_ugr_strand[rmdup_ugr_strand[, 2] == "-", 1])
strandInfo <- rep("*", length(gr))
strandInfo[setdiff(txdb_plus_gr_ind, revCompInd)] <- "+"
strandInfo[intersect(txdb_plus_gr_ind, revCompInd)] <- "-"
strandInfo[setdiff(txdb_minus_gr_ind, revCompInd)] <- "-"
strandInfo[intersect(txdb_minus_gr_ind, revCompInd)] <- "+"
warning("Out of ", length(context), " mutations, we could obtain",
"transcription direction information for ",
length(txdb_plus_gr_ind) + length(txdb_minus_gr_ind),
" mutation. Other mutations are removed.")
context <- context[strandInfo != "*"]
ref_base <- ref_base[strandInfo != "*"]
alt_base <- alt_base[strandInfo != "*"]
sampleName_str <- sampleName_str[strandInfo != "*"]
chrInfo <- chrInfo[strandInfo != "*"]
posInfo <- posInfo[strandInfo != "*"]
strandInfo <- strandInfo[strandInfo != "*"]
}
if (trDir == FALSE) {
strandInfo <- NULL
}
mutFeatures <- getMutationFeatureVector(context, ref_base, alt_base,
strandInfo, numBases, type)
suSampleStr <- sort(unique(sampleName_str))
lookupSampleInd <- seq_len(length(suSampleStr))
names(lookupSampleInd) <- suSampleStr
sampleIDs <- lookupSampleInd[sampleName_str]
featStr <- apply(mutFeatures, 1, paste0, collapse=",")
suFeatStr <- sort(unique(featStr))
lookupFeatInd <- seq_len(length(suFeatStr))
names(lookupFeatInd) <- suFeatStr
rawCount <- data.frame(sample = sampleIDs, mutInds = lookupFeatInd[featStr])
tableCount <- table(rawCount)
w <- which(tableCount > 0, arr.ind=TRUE)
procCount <- cbind(w[,2], w[,1], tableCount[w])
if (length(fdim) == 1) {
mutFeatList <- matrix(as.integer(suFeatStr), length(suFeatStr), 1)
} else {
mutFeatList <- t(vapply(suFeatStr,
function(x) as.numeric(unlist(strsplit(x, ","))),
numeric(numBases + as.integer(trDir))))
}
rownames(mutFeatList) <- NULL
rownames(procCount) <- NULL
if (trDir == FALSE) {
strandInfo_for_class <- rep(NA, length(chrInfo))
} else {
strandInfo_for_class <- strandInfo
}
return(methods::new(Class = "MutationFeatureData",
type = type,
flankingBasesNum = as.integer(numBases),
transcriptionDirection = trDir,
possibleFeatures = as.integer(fdim),
featureVectorList = t(mutFeatList),
sampleList = suSampleStr,
countData = t(procCount),
mutationPosition =
data.frame(chr = chrInfo, pos = posInfo,
ref = ref_base, alt = alt_base,
strand = strandInfo_for_class,
context = context,
sampleID = unname(sampleIDs),
mutID = unname(lookupFeatInd[featStr]),
stringsAsFactors = FALSE)
)
)
}
#' Get mutation feature vector from context sequence data and reference and
#' alternate allele information
#'
#' @param context the context sequence data around the mutated position. This
#' shoud be Biostrings::DNAStringSet class
#' @param ref_base the reference bases at the mutated position.
#' @param alt_base the alternate bases at the mutated position.
#' @param strandInfo transcribed strand information at the mutated position.
#' (this is optional)
#' @param numBases the number of flanking bases around the mutated position.
#' @param type the type of mutation feature vecotr (should be "independent" or
#' "full").
#' @return a mutation featuer vector
#'
#'
#'
getMutationFeatureVector <- function(context, ref_base, alt_base,
strandInfo = NULL, numBases, type) {
trDir <- !is.null(strandInfo)
if (type == "independent") {
fdim <- c(6, rep(4, numBases - 1), rep(2, as.integer(trDir)))
} else if (type == "full") {
fdim <- c(6 * 4^(numBases - 1) * 2^(as.integer(trDir)))
} else {
stop('the type argument has to be "independent" or "full"')
}
if (numBases %% 2 != 1) {
stop("numBases should be odd numbers")
}
centerInd <- (numBases + 1) / 2
if (type == "independent") {
mutFeatures <- matrix(0, length(ref_base), length(fdim))
centralBaseList <- data.frame(ref_base=rep(c("C", "T"), each=3),
alt_base=c("A", "G", "T", "A", "C", "G"),
stringsAsFactors = FALSE)
## rewrite the part from pmsignature
for (i in seq_len(6)) {
mutFeatures[which(ref_base == centralBaseList$ref_base[i] &
alt_base == centralBaseList$alt_base[i]), 1] <- i
}
columnInd <- 2
for (baseInd in seq_len(numBases)) {
if (baseInd == centerInd) {
next
}
baseList <- c("A", "C", "G", "T")
## rewrite the part from pmsignature
for (base in seq_along(baseList)) {
mutFeatures[which(XVector::subseq(context,
start=baseInd,
end=baseInd) == baseList[base]),
columnInd] <- base
}
columnInd <- columnInd + 1
}
if (trDir ==TRUE) {
mutFeatures[which(strandInfo == "+"), length(fdim)] <- 1
mutFeatures[which(strandInfo == "-"), length(fdim)] <- 2
}
} else {
mutFeatures <- matrix(1, length(ref_base), length(fdim))
tempDigits <- 1
for (i in seq_len((numBases - 1) / 2)) {
baseInd <- numBases + 1 - i
mutFeaturesList <- c("C", "G", "T")
for(i in mutFeaturesList) {
mutFeatures[which(XVector::subseq(context, start=baseInd, end=baseInd)
== i), 1] <-
mutFeatures[which(XVector::subseq(context, start=baseInd, end=baseInd)
== i), 1] + tempDigits * which(mutFeaturesList == i)
}
tempDigits <- tempDigits * 4
baseInd <- i;
for(i in mutFeaturesList) {
mutFeatures[which(XVector::subseq(context, start=baseInd, end=baseInd)
== i), 1] <-
mutFeatures[which(XVector::subseq(context, tart=baseInd, end=baseInd)
== i), 1] + tempDigits * which(mutFeaturesList == i)
}
tempDigits <- tempDigits * 4
}
mutFeaturesMat <- cbind(c("C", "C", "T", "T", "T"),
c("G", "T", "A", "C", "G"))
for(i in seq_len(nrow(mutFeaturesMat))) {
mutFea1 <- mutFeaturesMat[i, 1]
mutFea2 <- mutFeaturesMat[i, 2]
mutFeatures[which(ref_base == mutFea1 & alt_base == mutFea2), 1] <-
mutFeatures[which(ref_base == mutFea1 & alt_base == mutFea2), 1] +
tempDigits * i
}
if (trDir ==TRUE) {
tempDigits <- tempDigits * 6
mutFeatures[which(strandInfo == "-"), 1] <-
mutFeatures[which(strandInfo == "-"), 1] + tempDigits * 1
}
}
return(mutFeatures)
}
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.