Nothing
#' Import results from a STAR-Fusion run into a list of Fusion objects.
#'
#' A function that imports the results from a STAR-Fusion run, typically from
#' a star-fusion.fusion_candidates.final.abridged file, into a list of Fusion
#' objects.
#'
#' @param filename Filename for the STAR-Fusion
#' star-fusion.fusion_candidates.final.abridged results file.
#' @param genome_version Which genome was used in mapping (hg19, hg38, etc.).
#' @param limit A limit on how many lines to read.
#'
#' @return A list of Fusion objects.
#'
#' @examples
#' starfusionData <- system.file(
#' "extdata",
#' "star-fusion.fusion_candidates.final.abridged.txt",
#' package = "chimeraviz")
#' fusions <- import_starfusion(starfusionData, "hg19", 3)
#' # This should import a list of 3 fusions described in Fusion objects.
#'
#' @importFrom data.table fread
#'
#' @export
import_starfusion <- function (filename, genome_version, limit) {
# Is the genome version valid?
valid_genomes <- c("hg19", "hg38", "mm10")
if (is.na(match(tolower(genome_version), tolower(valid_genomes)))) {
stop("Invalid genome version given")
}
# If the limit is set, is the value valid?
if (missing(limit) == FALSE) {
if (is.numeric(limit) == FALSE || limit <= 0) {
stop("limit must be a numeric value bigger than 0")
}
}
# Try to read the fusion report
report <- withCallingHandlers({
col_types <- c(
"JunctionReadCount" = "integer",
"SpanningFragCount" = "integer",
"LeftGene" = "character",
"LeftBreakpoint" = "character",
"RightGene" = "character",
"RightBreakpoint" = "character",
"LargeAnchorSupport" = "character",
"LeftBreakDinuc" = "character",
"LeftBreakEntropy" = "numeric",
"RightBreakDinuc" = "character",
"RightBreakEntropy" = "numeric",
"FFPM" = "numeric"
)
if (missing(limit)) {
# Read all lines
data.table::fread(
input = filename,
colClasses = col_types,
showProgress = FALSE
)
} else {
# Only read up to the limit
data.table::fread(
input = filename,
colClasses = col_types,
showProgress = FALSE,
nrows = limit
)
}
},
error = function(cond) {
message(paste0("Reading ", filename, " caused an error: ", cond[[1]]))
stop(cond)
},
warning = function(cond) {
# Begin Exclude Linting
message(paste0("Reading ", filename, " caused a warning: ", cond[[1]]))
# End Exclude Linting
}
)
# Set variables
id <- NA
inframe <- NA
fusion_tool <- "starfusion"
spanning_reads_count <- NA
split_reads_count <- NA
# List to hold all Fusion objects
fusion_list <- vector("list", dim(report)[1])
# Iterate through each line in the .tsv file
for (i in 1:dim(report)[1]) {
# Import starfusion-specific fields
fusion_tool_specific_data <- list()
fusion_tool_specific_data[["LargeAnchorSupport"]] <-
report[[i, "LargeAnchorSupport"]]
fusion_tool_specific_data[["LeftBreakDinuc"]] <-
report[[i, "LeftBreakDinuc"]]
fusion_tool_specific_data[["LeftBreakEntropy"]] <-
report[[i, "LeftBreakEntropy"]]
fusion_tool_specific_data[["RightBreakDinuc"]] <-
report[[i, "RightBreakDinuc"]]
fusion_tool_specific_data[["RightBreakEntropy"]] <-
report[[i, "RightBreakEntropy"]]
fusion_tool_specific_data[["FFPM"]] <-
report[[i, "FFPM"]]
# Optional extra fields:
if(!is.null(report[[i, "PROT_FUSION_TYPE"]])) {
fusion_tool_specific_data[["PROT_FUSION_TYPE"]] <- report[[i, "PROT_FUSION_TYPE"]]
if (report[[i, "PROT_FUSION_TYPE"]] == 'INFRAME') {
inframe <- TRUE
} else if (report[[i, "PROT_FUSION_TYPE"]] == 'FRAMESHIFT') {
inframe <- FALSE
}
}
if(!is.null(report[[i, "annots"]])) {
fusion_tool_specific_data[["annots"]] <- report[[i, "annots"]]
}
# id for this fusion
id <- as.character(i)
left_breakpoint <-
unlist(strsplit(report[[i, "LeftBreakpoint"]], split = ":"))
righ_breakpoint <-
unlist(strsplit(report[[i, "RightBreakpoint"]], split = ":"))
# Chromosome names
chromosome_upstream <- left_breakpoint[1]
chromosome_downstream <- righ_breakpoint[1]
# Breakpoints
breakpoint_upstream <- as.numeric(left_breakpoint[2])
breakpoint_downstream <- as.numeric(righ_breakpoint[2])
# Strand
strand_upstream <- left_breakpoint[3]
strand_downstream <- righ_breakpoint[3]
# Number of supporting reads
split_reads_count <- report[[i, "JunctionReadCount"]]
spanning_reads_count <- report[[i, "SpanningFragCount"]]
# Fusion sequence. Ref https://github.com/stianlagstad/chimeraviz/pull/23
if(!is.null(report[[i, "FUSION_CDS"]])) {
junction_sequence_upstream <- Biostrings::DNAString(
gsub("[A-Z]+$", "", report[[i, "FUSION_CDS"]])
)
junction_sequence_downstream <- Biostrings::DNAString(
gsub("^[a-z]+", "", report[[i, "FUSION_CDS"]])
)
} else {
junction_sequence_upstream <- Biostrings::DNAString()
junction_sequence_downstream <- Biostrings::DNAString()
}
# Gene names
gene_names_1 <- unlist(strsplit(report[[i, "LeftGene"]], split = "\\^"))
gene_names_2 <- unlist(strsplit(report[[i, "RightGene"]], split = "\\^"))
name_upstream <- gene_names_1[1]
name_downstream <- gene_names_2[1]
# Ensembl ids
ensembl_id_upstream <- strsplit(gene_names_1[2], "\\.")[[1]][[1]]
ensembl_id_downstream <- strsplit(gene_names_2[2], "\\.")[[1]][[1]]
# PartnerGene objects
gene_upstream <- new(
Class = "PartnerGene",
name = name_upstream,
ensembl_id = ensembl_id_upstream,
chromosome = chromosome_upstream,
breakpoint = breakpoint_upstream,
strand = strand_upstream,
junction_sequence = junction_sequence_upstream,
transcripts = GenomicRanges::GRangesList()
)
gene_downstream <- new(
Class = "PartnerGene",
name = name_downstream,
ensembl_id = ensembl_id_downstream,
chromosome = chromosome_downstream,
breakpoint = breakpoint_downstream,
strand = strand_downstream,
junction_sequence = junction_sequence_downstream,
transcripts = GenomicRanges::GRangesList()
)
fusion_list[[i]] <- new(
Class = "Fusion",
id = id,
fusion_tool = fusion_tool,
genome_version = genome_version,
spanning_reads_count = spanning_reads_count,
split_reads_count = split_reads_count,
fusion_reads_alignment = Gviz::AlignmentsTrack(),
gene_upstream = gene_upstream,
gene_downstream = gene_downstream,
inframe = inframe,
fusion_tool_specific_data = fusion_tool_specific_data
)
}
# Return the list of Fusion objects
fusion_list
}
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.