R/plotCNA.R

In CopywriteR: Copy number information from targeted sequencing using off-target reads

plotCNA <- function(destination.folder, smoothed = TRUE, sample.plot, y.min,
                    y.max, ...) {

    start.time <- Sys.time()
    
    ## Restore work directory upon exit
    wd.orig <- getwd()
    on.exit(setwd(wd.orig))

    ## Make destination folder path absolute
    destination.folder <- tools::file_path_as_absolute(destination.folder)
    destination.folder <- file.path(destination.folder, "CNAprofiles")

    ## Provide output to log
    flog.appender(appender.file(file.path(destination.folder,
                                          "CopywriteR.log")))
    args.ellipsis <- unlist(list(...))
    flog.info(paste0("plotCNA was run using the following commands:", "\n\n",
                     "plotCNA(destination.folder = \"",
                     dirname(destination.folder),
                     "\", smoothed = ", smoothed,
                     if (!missing(sample.plot)) {", sample.plot = sample.plot"},
                     if (!missing(y.min)) {paste0(", y.min = ", y.min)},
                     if (!missing(y.max)) {paste0(", y.max = ", y.max)},
                     if (!missing(...)) {paste0(", ", paste(names(args.ellipsis),
                                                "=", args.ellipsis,
                                                collapse = ", "))},
                     ")"))

    ## Check the existence of folders and files
    if (!file.exists(destination.folder)) {
        stop(.wrap("The destination folder could not be found. Please change",
                   "the path specified in", sQuote(destination.folder)))
    }

    # Pre-define inputStructure variable to avoid it from raising NOTES during
    # R CMD CHECK (variable is loaded from file below)
    inputStructure <- NULL

    # Load inputStructure variable
    load(file.path(destination.folder, "input.Rdata"))

    ## Set variables
    chromosomes <- inputStructure$chromosomes
    nautosomes <- length(grep("[0-9]", chromosomes))
    prefix <- inputStructure$prefix
    bin.size <- inputStructure$bin.size

    ## Set sample.plot
    if (missing(sample.plot)) {
        sample.plot <- apply(inputStructure$sample.control, c(1, 2),
                             function(x) {
            x <- paste0("log2.", basename(x))
        })
        all.samples <- unique(as.vector(sample.plot))
        sample.plot <- rbind(data.frame(sample.plot, stringsAsFactors = FALSE),
                             data.frame(samples = all.samples,
                                        controls = rep(NA, length(all.samples)), 
                                        stringsAsFactors = FALSE))
    } else {
        colnames(sample.plot) <- c("samples", "controls")
        sample.plot[, ] <- apply(sample.plot, c(1, 2), function(x) {
            if (!is.na(x)) {
                x <- paste0("log2.", gsub("_properreads", "", basename(x)))
            } else {
                x <- as.character(x)
            }
        })
    }

    ## Read data
    log2.read.counts <- read.table(file = file.path(destination.folder,
                                                    "log2_read_counts.igv"),
                                   sep = "\t", header = TRUE,
                                   check.names = FALSE)

    ## Remove prefix and convert chromosome names to integers
    log2.read.counts$Chromosome <- gsub(prefix, "", log2.read.counts$Chromosome)
    chromosomes <- gsub(prefix, "", chromosomes)
    
    log2.read.counts$Chromosome <- gsub("X", nautosomes + 1,
                                        log2.read.counts$Chromosome)
    chromosomes <- gsub("X", nautosomes + 1, chromosomes)
    
    log2.read.counts$Chromosome <- gsub("Y", nautosomes + 2,
                                        log2.read.counts$Chromosome)
    chromosomes <- gsub("Y", nautosomes + 2, chromosomes)
    
    log2.read.counts$Chromosome <- as.integer(log2.read.counts$Chromosome)
    chromosomes <- sort(as.integer(chromosomes))

    ## Fix behaviour of DNAcopy with 'outlier' values
    log2.read.counts[, 5:ncol(log2.read.counts)] <-
        apply(log2.read.counts[, 5:ncol(log2.read.counts), drop = FALSE],
              c(1, 2), function(x) {
        if (is.na(x)) {
            x <- NA
        } else if (x < -5) {
            x <- -5
        } else if (x > 10) {
            x <- 10
        } else {
            x <- x
        }
    })

    ## Create table with values to be plotted
    if (all(na.omit(unlist(sample.plot)) %in% colnames(log2.read.counts))) {
        plotting.values <-
            log2.read.counts[, c("Chromosome", "Start", "End", "Feature")]
        for (i in seq_len(nrow(sample.plot))) {
            if (!is.na(sample.plot$controls[i])) {
                plotting.values[, paste0(sample.plot$samples[i], ".vs.",
                                         sample.plot$controls[i])] <- 
                    log2.read.counts[, sample.plot$samples[i]] -
                    log2.read.counts[, sample.plot$controls[i]]
            } else {
                plotting.values[, paste0(sample.plot$samples[i], ".vs.none")] <-
                    log2.read.counts[, sample.plot$samples[i]]
            }
        }
    } else {
        unique.samples <- unique(na.omit(unlist(sample.plot)))
        print(unique.samples)
        print(unique.samples %in% colnames(log2.read.counts))
        missing.samples <- unique.samples[!unique.samples %in% colnames(log2.read.counts)]
        print(missing.samples)
        stop(.wrap("The following samples refer to .bam files that have not",
                   "been processed by CopywriteR:",
                   paste(sQuote(missing.samples), collapse = ", "),
                   ". Please re-analyze all required samples using CopywriteR",
                   "and run the plotCNA function again."))
    }

    ## Apply DNAcopy
    CNA.object <- CNA(plotting.values[, 5:ncol(plotting.values), drop = FALSE],
                      plotting.values$Chromosome,
                      rowMeans(plotting.values[, c("Start", "End")]),
                      data.type = "logratio", 
                      sampleid = colnames(plotting.values)[5:ncol(plotting.values)])
    if (smoothed) {
        CNA.object <- smooth.CNA(CNA.object)
    }
    segment.CNA.object <- segment(CNA.object, verbose = 1, ...)
    save(segment.CNA.object, file = file.path(destination.folder,
                                              "segment.Rdata"))

    ## Calculate the chromosome lengths from the bin.bed file
    chrom.lengths <-
        scanBamHeader(inputStructure$sample.control$samples[1])[[1]]$targets
    names(chrom.lengths) <- gsub(prefix, "", names(chrom.lengths))    
    chrom.lengths <- data.frame(Chromosome = names(chrom.lengths),
                                Length = chrom.lengths, row.names = NULL,
                                stringsAsFactors = FALSE)
    suppressWarnings(chrom.lengths <- chrom.lengths[chrom.lengths$Chromosome %in% c("X", "Y") |
                                                    !is.na(as.integer(as.vector(chrom.lengths$Chromosome))), ])
    chrom.lengths <- chrom.lengths[mixedorder(chrom.lengths$Chromosome), ]
    chrom.lengths$Chromosome <- gsub("X", nautosomes + 1,
                                     chrom.lengths$Chromosome)
    chrom.lengths$Chromosome <- gsub("Y", nautosomes + 2,
                                     chrom.lengths$Chromosome)
    chrom.lengths[, "CumSum"] <- c(0, cumsum(as.numeric(chrom.lengths$Length))[seq_len(nrow(chrom.lengths) - 1)])
    
    ## Create plots
    segment.CNA.object$output[, "start.position.chrom"] <-
        chrom.lengths$CumSum[match(as.integer(segment.CNA.object$output$chrom),
                                   chrom.lengths$Chromosome)]
    segment.CNA.object$output$loc.start <- segment.CNA.object$output$loc.start + segment.CNA.object$output$start.position.chrom
    segment.CNA.object$output$loc.end <- segment.CNA.object$output$loc.end + segment.CNA.object$output$start.position.chrom
    segment.CNA.object$output$start.position.chrom <- NULL
    
    segment.CNA.object$data[, "start.position.chrom"] <-
        chrom.lengths$CumSum[match(as.integer(segment.CNA.object$data$chrom),
                                   chrom.lengths$Chromosome)]
    segment.CNA.object$data$maploc <- segment.CNA.object$data$maploc + segment.CNA.object$data$start.position.chrom
    segment.CNA.object$data$start.position.chrom <- NULL

    # Get sample names
    samples <- colnames(segment.CNA.object$data)
    samples <- samples[3:length(samples)]

    # Create plots folder
    plot.folder <- file.path(destination.folder, "plots")
    dir.create(plot.folder)
    setwd(plot.folder)

    # Set boundaries plots
    if (missing(y.min)) {
        y.min <- -2
    }
    if (missing(y.max)) {
        y.max <- 5
    }

    # Loop through samples using lapply
    invisible(lapply(samples, function(x) {
        # Select sample
        select.sample <- x
        current.sample <- segment.CNA.object
        current.sample$data <-
            current.sample$data[, c("chrom", "maploc", select.sample)]
        current.sample$data <-
            current.sample$data[!is.na(current.sample$data[, select.sample]), ]
        current.sample$output <-
            current.sample$output[current.sample$output$ID == select.sample, ]

        # Print MAD-values in log file
        flog.info(paste("The MAD-value for sample", select.sample, "is:",
                        round(madDiff(current.sample$data[, select.sample]), 3)))
                            
        # Create and set new directory
        dir.create(file.path(plot.folder, select.sample))
        setwd(file.path(plot.folder, select.sample))

        # Loop through chromosomes using lapply
        invisible(lapply(c(as.list(chromosomes), list(chromosomes)),
                         function(x) {
            # Select chromosome
            select.chrom <- x
            current.sample$data <- current.sample$data[current.sample$data$chrom %in% select.chrom, ]
            current.sample$output <- current.sample$output[current.sample$output$chrom %in% select.chrom, ]

            # Set variables
            genome.position.min <-
                chrom.lengths$CumSum[match(min(select.chrom),
                                           chrom.lengths$Chromosome)]
            genome.position.max <-
                chrom.lengths$CumSum[match(max(select.chrom),
                                           chrom.lengths$Chromosome)] +
                                     chrom.lengths$Length[match(max(select.chrom), chrom.lengths$Chromosome)]

            # Plot data
            if (length(x) > 1) {
                name.chrom <- "all_chrom"
            } else {
                name.chrom <- paste0("chrom_", x)
            }
            pdf(file = paste0(name.chrom, ".pdf"), width = 14, height = 7)
            plot(current.sample$data[, "maploc"],
                 current.sample$data[, select.sample], ylim = c(y.min, y.max),
                 pch = ".", cex = 1, xlim = c(genome.position.min, 
                                              genome.position.max),
                 xaxs = "i", xlab = "", ylab = "log2 value", xaxt = "n",
                 main = select.sample)

            points(current.sample$data$maploc[current.sample$data[, select.sample] > y.max],
                   rep.int(y.max, sum(current.sample$data[, select.sample] > y.max)),
                   cex = 0.25, col = "green", pch = 2)
            points(current.sample$data$maploc[current.sample$data[, select.sample] < y.min],
                   rep.int(y.min, sum(current.sample$data[, select.sample] < y.min)),
                   cex = 0.25, col = "red", pch = 6)

            invisible(apply(current.sample$output, 1, function(x) {
                segments(x0 = as.numeric(x["loc.start"]),
                         y0 = as.numeric(x["seg.mean"]),
                         x1 = as.numeric(x["loc.end"]),
                         col = "red", pch = ".", cex = 0.2)
            }))
            par(xpd = TRUE)
            text(x = 0.96 * (genome.position.max - genome.position.min) + genome.position.min,
                 y = 0.075 * (y.max - y.min) + y.max,
                 labels = paste0("mad = ", round(madDiff(current.sample$data[, select.sample]), 3)))
            text(x = 0.035 * (genome.position.max - genome.position.min) + genome.position.min,
                 y = 0.075 * (y.max - y.min) + y.max,
                 labels = paste0(bin.size / 1000, " kb bins"))
            par(xpd = FALSE)
            ticks <- (chrom.lengths$CumSum + chrom.lengths$Length / 2)[select.chrom]
            axis(1, at = ticks, labels = select.chrom)
            if (length(select.chrom) > 1) {
                abline(v = chrom.lengths$CumSum[2:nrow(chrom.lengths)],
                       lty = "dotted")
            }
            dev.off()
        }))
    }))
    flog.info(paste("Total calculation time of plotCNA was",
                    round(difftime(Sys.time(), start.time, units = "mins"), 2),
                    "minutes"))
    cat("Total calculation time of CopywriteR was: ",
        Sys.time() - start.time, "\n\n")
}