R/builds.R

In waldronlab/TCGAmisc: TCGA utility functions for data management

human_builds <- function() {
    S4Vectors::DataFrame(
        Date = c("July 2004", "May 2004", "March 2006", "February 2009",
            "December 2013"),
        NCBI_PRE = c("NCBI", "NCBI", "NCBI", "GRCh", "GRCh"),
        NCBI_NO = c("34", "35", "36", "37", "38"),
        NCBI =  c("NCBI34", "NCBI35", "NCBI36", "GRCh37", "GRCh38"),
        UCSC_PRE = c("hg", "hg", "hg", "hg", "hg"),
        UCSC_NO = c("16", "17", "18", "19", "38"),
        UCSC = c("hg16", "hg17", "hg18", "hg19", "hg38")
    )
}

#' @name builds
#'
#' @title Utilities for working with *HUMAN* genome builds
#'
#' @description A few functions are available to search for build versions,
#' either from NCBI or UCSC.
#'
#' \itemize{
#'   \item `translateBuild`: translates between UCSC and NCBI build
#'   versions
#'   \item `extractBuild`: use grep patterns to find the first build
#'   within the string input
#'   \item `uniformBuilds`: replace build occurrences below a threshold
#'   level of occurence with the alternative build
#'   \item `correctBuild`: Ensure that the build annotation is correct
#'   based on the NCBI/UCSC website. If not, use `translateBuild` with
#'   the indicated 'style' input
#'   \item `isCorrect`: Check to see if the build is exactly as annotated
#' }
#'
#' @details The `correctBuild` function takes the input and ensures that
#' the style specified matches the input. Otherwise, it will
#' return the correct style for use with  `seqlevelsStyle`.
#' Currently, the function does not support patched builds
#' (e.g., 'GRCh38.p13') Build names are taken from the website:
#' \url{https://www.ncbi.nlm.nih.gov/assembly/GCF_000001405.26/}
#'
#' @param from character() A vector of build versions typically from `genome()`
#'     (e.g., "37"). The build vector must be homogenous (i.e.,
#'     `length(unique(x)) == 1L`).
#'
#' @param to character(1) The name of the desired build version (either "UCSC"
#'     or "NCBI"; default: "UCSC")
#'
#' @param build character(1) A string providing the genome build
#'
#' @param style character(1) The annotation style, either 'UCSC' or 'NCBI'
#'
#' @examples
#'
#' translateBuild("GRCh35", "UCSC")
#'
#' @return
#'     translateBuild: A character vector of translated genome builds
#'
#'     extractBuild: A character string of the build information available
#'
#'     uniformBuilds: A character vector of builds where all builds are
#'         identical `identical(length(unique(build)), 1L)`
#'
#'     correctBuild: A character string of the 'corrected' build name
#'
#'     isCorrect: A logical indicating if the build is exactly as annotated
#'
#' @export
translateBuild <- function(from, to = c("UCSC", "NCBI")) {
    lfro <- length(from)
    from <- unique(from)
    if (!.isSingleValue(from))
        stop("Enter a consistent vector of genomic builds")

    to <- match.arg(to)
    buildDF <- human_builds()

    bnames <- c("UCSC", "NCBI")
    from_build <- bnames[bnames != to]

    bfrom <- correctBuild(from, from_build)

    buildIndex <- match(bfrom, buildDF[[from_build]])
    rep(buildDF[[to]][buildIndex], lfro)
}

#' @rdname builds
#'
#' @param build character(1) A string providing the genome build
#'
#' @param style character(1) The annotation style, either 'UCSC' or 'NCBI'
#'
#' @examples
#'
#' correctBuild("grch38", "NCBI")
#' correctBuild("hg19", "NCBI")
#'
#' @export
correctBuild <- function(build, style = c("UCSC", "NCBI")) {
    build.df <- human_builds()
    pre <- paste0(style, "_PRE")
    digits <- as.character(gsub(".*([[:digit:]]{2})", "\\1", build))
    pref <- gsub("(.*)([[:digit:]]{2})", "\\1", build)
    if (identical(tolower(pref), "hg") && identical(style, "NCBI"))
        return(translateBuild(build, style))
    if (
        tolower(pref) %in% tolower(build.df[["NCBI_PRE"]]) &&
        identical(style, "UCSC")
    )
        return(translateBuild(build, style))
    idx <- match(digits, build.df[[paste0(style, "_NO")]])
    if (is.na(idx))
        return(NA_character_)
    num <- build.df[[paste0(style, "_NO")]][idx]
    pref <- build.df[[pre]][idx]
    paste0(pref, num)
}

#' @rdname builds
#'
#' @examples
#'
#' isCorrect("GRCh38", "NCBI")
#'
#' isCorrect("hg19", "UCSC")
#'
#' @export
isCorrect <- function(build, style = c("UCSC", "NCBI")) {
    identical(
        correctBuild(build, style),
        build
    )
}

#' @rdname builds
#'
#' @param string A single character string
#' @param build A vector of build version names (default UCSC, NCBI)
#'
#' @examples
#'
#' extractBuild(
#' "SCENA_p_TCGAb29and30_SNP_N_GenomeWideSNP_6_G05_569110.nocnv_grch38.seg.txt"
#' )
#'
#' @export
extractBuild <- function(string, build = c("UCSC", "NCBI")) {
    if (!S4Vectors::isSingleString(string))
        stop("Provide a single string for build search")
    builds <- vector(mode = "character", length(build))
    names(builds) <- build
    for (i in build) {
        pattrn <- switch(i, UCSC = "[Hh][Gg][0-9]{2}",
            NCBI = "[Gg][Rr][Cc][Hh][0-9]{2}")
        builds[[i]] <- stringr::str_extract(string, pattrn)
    }
    builds <- Filter(function(x) !is.na(x), builds)
    if (!length(builds))
        NA_character_
    else if (length(builds))
        builds[1L]
}

.isSingleValue <- function(charvec) {
    identical(length(unique(charvec)), 1L)
}

.consistentNumbers <- function(charvec) {
    bnos <- gsub("(.*)([0-9]{2})", "\\2", charvec)
    .isSingleValue(bnos)
}


.replaceHighProp <- function(charvec) {
    tt <- table(charvec)
    if (length(tt) > 2L)
        stop("<internal> Table has more than 2 values")

    proptt <- prop.table(tt)

    highprop <- names(which.max(proptt))
    charvec[charvec != highprop] <- highprop
    charvec
}

#' @rdname builds
#'
#' @param builds A character vector of builds
#'
#' @param cutoff numeric(1L) An inclusive threshold tolerance value for missing
#'     values and translating builds that are below the threshold
#'
#' @param na character() The values to be considered as missing (default:
#'     c("", "NA"))
#'
#' @examples
#'
#' buildvec <- rep(c("GRCh37", "hg19"), times = c(5, 1))
#' uniformBuilds(buildvec)
#'
#' navec <- c(rep(c("GRCh37", "hg19"), times = c(5, 1)), "NA")
#' uniformBuilds(navec)
#'
#' @export uniformBuilds
uniformBuilds <- function(builds, cutoff = 0.2, na = c("", "NA")) {
    tbuild <- table(builds)
    if (.consistentNumbers(builds)) {
        if (identical(length(tbuild), 1L))
            return(builds)
        else
            builds <- .replaceHighProp(builds)
    }

    wbuilds <- toupper(builds)
    nabuilds <- wbuilds %in% na | is.na(wbuilds)
    wbuilds[nabuilds] <- NA_character_

    tt <- table(wbuilds, useNA = "always")
    proptt <- prop.table(tt)

    uvals <- names(proptt)
    nanames <- is.na(uvals)
    propna <- proptt[nanames]

    if (propna >= cutoff)
        stop("Frequency of NA values higher than the cutoff tolerance")

    ubuilds <- uvals[!nanames]

    if (.isSingleValue(ubuilds)) {
        builds[nabuilds] <- ubuilds
        return(builds)
    } else if (sum(!nanames) > 2)
        stop("Only two build types at a time can be used")

    props <- proptt[!nanames]

    offbuild <- names(props[props <= cutoff])
    mainbuild <- names(props[props > cutoff])
    mainbuild <- builds[match(mainbuild, toupper(builds))]
    if (any(nabuilds))
        builds[nabuilds] <- mainbuild

    samebuilds <- .consistentNumbers(builds)
    if (samebuilds) {
        builds[wbuilds == offbuild] <- mainbuild
    } else {
        pattrn <- vapply(
            c(UCSC = "[Hh][Gg][0-9]{2}", NCBI = "[Gg][Rr][Cc][Hh][0-9]{2}"),
            grepl, logical(1L), offbuild)
        toconv <- names(pattrn)[!pattrn]
        results <- translateBuild(offbuild, toconv)
        builds[wbuilds == offbuild] <- results
    }
    builds
}