R/get_sv.R

In sigminer: Extract, Analyze and Visualize Mutational Signatures for Genomic Variations

# Handling RS classes -------------------------------------------------------

#' Read Structural Variation Data as RS object
#'
#' @param input a `data.frame` or a file with the following columns:
#' "sample", "chr1", "start1", "end1", "chr2", "start2", "end2", "strand1", "strand2", "svclass".
#' NOTE: If column "svclass" already exists in input, "strand1" and "strand2" are optional.
#' If "svclass" is not provided, `read_sv_as_rs()` will compute it by
#' "strand1","strand2"(strand1/strand2),"chr1" and "chr2":
#' - translocation, if mates are on different chromosomes.
#' - inversion (+/-) and (-/+), if mates on the same chromosome.
#' - deletion (+/+), if mates on the same chromosome.
#' - tandem-duplication (-/-), if mates on the same chromosome.
#' @return a `list`
#' @export
#'
#' @examples
#' sv <- readRDS(system.file("extdata", "toy_sv.rds", package = "sigminer", mustWork = TRUE))
#' rs <- read_sv_as_rs(sv)
#' # svclass is optional
#' rs2 <- read_sv_as_rs(sv[, setdiff(colnames(sv), "svclass")])
#' identical(rs, rs2)
#' \dontrun{
#' tally_rs <- sig_tally(rs)
#' }
#' @testexamples
#' expect_is(rs, "RS")
#' expect_is(tally_rs, "list")
read_sv_as_rs <- function(input) {
  if (is.data.frame(x = input)) {
    input <- data.table::as.data.table(input)
  } else {
    input <- data.table::fread(
      file = input,
      data.table = TRUE, header = TRUE
    )
  }

  # check necessary name
  necessary.fields <- c(
    "sample",
    "chr1", "start1", "end1",
    "chr2", "start2", "end2"
  )
  colnames(input) <- tolower(colnames(input))

  idx <- necessary.fields %in% colnames(input)
  if (!all(idx)) {
    stop(
      "Missing required columns from input: ",
      paste(necessary.fields[!idx], collapse = ", ")
    )
  }

  # check svclass present or not.
  # If not, generate it.
  if (!"svclass" %in% colnames(input)) {
    if (all(c("strand1", "strand2") %in% colnames(input))) {
      input <- get_svclass(input)
    } else {
      stop(
        "Can't computate because of missing additional columns: 'strand1' and 'strand2'"
      )
    }
  }

  # drop unnecessary fields
  input <- subset(input, select = c(necessary.fields, "svclass"))

  # chromosome "chr+number" to "number"
  input$chr1 <- sub("chr", "", input$chr1, ignore.case = TRUE)
  input$chr2 <- sub("chr", "", input$chr2, ignore.case = TRUE)

  class(input) <- c("RS", class(input))
  message("succesfully read RS!")
  return(input)
}

# split by sample and collect in a list: svlist --------------------------
get_svlist <- function(data) {
  index <- seq(1, nrow(data))
  data$Index <- index
  res <- split(data, by = "sample")
}

# get svclass
get_svclass <- function(data) {
  data %>% dplyr::mutate(
    svclass = dplyr::case_when(
      .data$chr1 != .data$chr2 ~ "translocation",
      .data$strand1 != .data$strand2 ~ "inversion",
      .data$strand1 == "-" & .data$strand2 == "-" ~ "tandem-duplication",
      .data$strand1 == "+" & .data$strand2 == "+" ~ "deletion"
    )
  )
}


# get size
getRearrSize <- function(sv_profiles) {
  rearrsize <- purrr::map_df(sv_profiles, function(x) {
    x %>%
      dplyr::mutate(
        length = x$end2 - x$start1,
        rearrsize = dplyr::case_when(
          .data$svclass == "translocation" ~ NA_character_,
          .data$length < 1000 ~ "<1Kb",
          .data$length >= 1000 & .data$length < 10000 ~ "1-10Kb",
          .data$length >= 10000 & .data$length < 100000 ~ "10-100Kb",
          .data$length >= 100000 & .data$length < 1000000 ~ "100Kb-1Mb",
          .data$length >= 1000000 & .data$length <= 10000000 ~ "1Mb-10Mb",
          .data$length > 10000000 ~ ">10Mb"
        )
      ) %>%
      dplyr::select_at(c("sample", "rearrsize", "Index"))
  })
  colnames(rearrsize) <- c("sample", "value", "Index")
  rearrsize <- rearrsize %>%
    data.table::as.data.table()
  rearrsize[order(rearrsize$Index)]
}

# copy from UCL-Research-Department-of-Pathology/RESIN --------------------
getDists <- function(chrom1, pos1, chrom2, pos2, doPCF = FALSE) {
  pos1 <- as.numeric(pos1)
  pos2 <- as.numeric(pos2)
  # get segments per chromosome
  dists <- sapply(unique(c(chrom1, chrom2)), FUN = function(x) {
    bp <- matrix(NA, ncol = 2, nrow = 0)
    # positions separate for each member of breakpoint
    pos <- c()
    index1 <- which(chrom1 == x)
    if (length(index1) > 0) {
      pos <- c(pos, pos1[index1])
      bp <- rbind(bp, cbind(index1, 1))
    }
    index2 <- which(chrom2 == x)
    if (length(index2) > 0) {
      pos <- c(pos, pos2[index2])
      bp <- rbind(bp, cbind(index2, 2))
    }
    # distances between breakpoint
    dists <- diff(sort(pos))
    # segment
    forCN <- data.frame(chrom = x, pos = sort(pos), dist = dists[order(pos)])
    if (!doPCF) {
      return(forCN[, 3])
    }
    pcf = "copynumber"%:::%"pcf"
    return(list(info = forCN, seg = pcf(forCN, gamma = 25, kmin = 10)))
  }, simplify = FALSE)
  return(dists)
}

# get clustered -----------------------------------------------------------

getClustered <- function(sv_profiles, threshold = NULL) {
  # each list each row apply function
  clustered <- purrr::map(sv_profiles, function(x) {
    # get pos and chrom
    pos1 <- x$start1
    pos2 <- x$start2
    chrom1 <- x$chr1
    chrom2 <- x$chr2

    # get segments per chromosome
    dists <- getDists(chrom1, pos1, chrom2, pos2, doPCF = TRUE)
    if (is.null(threshold)) threshold <- 0.1 * mean(unlist(sapply(dists, FUN = function(x) x$info[, 3])), na.rm = TRUE)

    # which segments are below threshold
    regions <- do.call(rbind, sapply(dists, FUN = function(x) {
      x$seg[which(x$seg$mean < threshold), ]
    }, simplify = FALSE))
    if (nrow(regions) == 0) {
      clustered <- rep("non-clustered", length(chrom1))
    } else {
      # which rearrangements are in clustered regions
      regionGR <- as(paste0(regions$chrom, ":", regions$start.pos, "-", regions$end.pos), "GRanges")
      clustered <- rep("non-clustered", length = length(pos1))
      index1 <- IRanges::findOverlaps(as(paste0(chrom1, ":", pos1, "-", pos1), "GRanges"), regionGR)@from
      if (length(index1) > 0) clustered[index1] <- "clustered"
      index2 <- IRanges::findOverlaps(as(paste0(chrom2, ":", pos2, "-", pos2), "GRanges"), regionGR)@from
      if (length(index2) > 0) clustered[index2] <- "clustered"
    }
    x$clustered <- clustered
    x[, c("sample", "clustered", "Index"), with = FALSE]
  })


  clustered_dt <- data.table::rbindlist(clustered)
  colnames(clustered_dt) <- c("sample", "value", "Index")
  clustered_dt[order(clustered_dt$Index)]
}


# get type ----------------------------------------------------------------
getType <- function(sv_profiles) {
  type_map <- c("del", "inv", "tds", "trans")
  names(type_map) <- c("deletion", "inversion", "tandem-duplication", "translocation")

  type <- purrr::map_df(sv_profiles, function(x) {
    x$type <- as.character(type_map[x$svclass])
    x[, c("sample", "type", "Index"), with = FALSE]
  })
  colnames(type) <- c("sample", "value", "Index")
  type <- type %>%
    data.table::as.data.table()
  type[order(type$Index)]
}


# get features list -------------------------------------------------------
get_features_sv <- function(sv_data) {

  field <- c(
    "clustered",
    "type",
    "size"
  )
  .get_feature <- function(i) {
    if (i == "clustered") {
      print("Getting clustered info...")
      zz <- getClustered(sv_data)
      zz
    }
    else if (i == "type") {
      print("Getting type of segment ...")
      getType(sv_data)
    }
    else if (i == "size") {
      print("Getting distance of two rearrange segments ...")
      getRearrSize(sv_data)
    }
  }
  res <- furrr::future_map(field, .get_feature,
    .progress = TRUE,
    .options = furrr::furrr_options(seed = TRUE)
  )
  res <- res %>% setNames(field)
  res
}

get_components_sv <- function(CN_features) {
  feature_names <- names(CN_features)
  purrr::map2(CN_features[feature_names], feature_names,
    .f = call_component_sv
  )
}

call_component_sv <- function(f_dt, f_name) {
  f_dt <- data.table::copy(f_dt)
  if (f_name == "clustered") {
    f_dt$C_clustered <- factor(f_dt$value, levels = c("clustered", "non-clustered"))
  }
  if (f_name == "type") {
    f_dt$C_type <- factor(f_dt$value, levels = c("del", "inv", "tds", "trans"))
  }
  if (f_name == "size") {
    f_dt$C_size <- factor(f_dt$value, levels = c("<1Kb", ">10Mb", "1-10Kb", "10-100Kb", "100Kb-1Mb", "1Mb-10Mb"))
  }
  f_dt$value <- NULL
  f_dt
}

# get matrix -------------------------------------------------------
get_matrix_sv <- function(CN_components, indices = NULL) {
  merged_dt <- purrr::reduce(CN_components, merge, by = c("sample", "Index"), all = TRUE)
  dt_mg <- merged_dt
  sv_class_levels <- vector_to_combination(
    levels(dt_mg$C_clustered),
    levels(dt_mg$C_type),
    levels(dt_mg$C_size),
    c_string = ":"
  ) %>%
    ifelse(grepl("trans", .),
      sub("trans.+", "trans", .),
      .
    ) %>%
    unique()

  dt_mg$sv_class <- paste(dt_mg$C_clustered, dt_mg$C_type, dt_mg$C_size, sep = ":") %>%
    ifelse(grepl("trans", .),
      sub("trans.+", "trans", .),
      .
    )
  dt_mg$sv_class <- factor(dt_mg$sv_class, levels = sv_class_levels)

  sv_mat <- classDT2Matrix(dt_mg, samp_col = "sample", component_col = "sv_class") %>%
    as.data.frame()

  sv_mat <- as.matrix(sv_mat[, sort(colnames(sv_mat))])

  sv_mat_32 <- sv_mat[, !(grepl("<1Kb", colnames(sv_mat)))]

  return(list(
    RS_32 = sv_mat_32,
    RS_38 = sv_mat
  ))
}