R/plotTranscripts-methods.R

In SplicingGraphs: Create, manipulate, visualize splicing graphs, and assign RNA-seq reads to them

### =========================================================================
### plotTranscripts()
### -------------------------------------------------------------------------

### Slighly faster and less memory consuming than subsetByOverlaps()
### when 'subject' contains 1 range only. Ignores the strand.
.subsetByOverlapWithSingleRange <- function(query, subject)
{
    if (!(is(query, "GAlignments") || is(query, "GAlignmentPairs")))
        stop("'query' must be a GAlignments or GAlignmentPairs object")
    if (!is(subject, "GenomicRanges"))
        stop("'subject' must be a GenomicRanges object")
    if (length(subject) != 1L)
        stop("'subject' must contain 1 range only")

    ## merge() will check that 'query' and 'subject' are based on compatible
    ## reference genomes.
    merge(seqinfo(query), seqinfo(subject))
    ## Drop all sequence levels but the one level that is used by the single
    ## range in 'subject'.
    seqlevels(subject) <- as.character(seqnames(subject))
    ## Set the same one sequence level on 'query'. Use 'pruning.mode="coarse"'
    ## to remove the elements in 'query' that are not on that sequence level.
    seqlevels(query, pruning.mode="coarse") <- seqlevels(subject)

    query_ranges <- granges(query)
    strand(query_ranges) <- "*"
    cmp <- pcompare(query_ranges, subject)
    ## Keep reads that overlap with or are adjacent to 'subject'.
    query[-5L <= cmp & cmp <= 5L]
}

.plotTranscripts.GRangesList <- function(x, reads=NULL,
                                         from=NA, to=NA, max.plot.reads=200)
{
    ## Check 'max.plot.reads'.
    if (!isSingleNumber(max.plot.reads))
        stop("'max.plot.reads' must be a single number")
    if (!is.integer(max.plot.reads))
        max.plot.reads <- as.integer(max.plot.reads)
    if (max.plot.reads < 0L)
        stop("'max.plot.reads' cannot be negative")

    ## Compute the genomic range of the transcripts.
    unlisted_x <- unlist(x, use.names=FALSE)
    strand(unlisted_x) <- "*"
    plot_range <- range(unlisted_x)
    if (length(plot_range) != 1L)
        stop("cannot plot transcripts that are on different chromosomes")

    ## Compute 'from' and 'to'.
    if (!(is.null(from) || isSingleNumberOrNA(from)))
        stop("'from' must be a single number, or NA, or NULL")
    if (!(is.null(to) || isSingleNumberOrNA(to)))
        stop("'to' must be a single number, or NA, or NULL")
    from_is_na <- !is.null(from) && is.na(from)
    to_is_na <- !is.null(to) && is.na(to)
    if (from_is_na || to_is_na) {
        x_min_start <- start(plot_range)
        x_max_end <- end(plot_range)
        margin <- 0.10 * (x_max_end - x_min_start)
        if (from_is_na)
            from <- x_min_start - margin
        if (to_is_na)
            to <- x_max_end + margin
        from_str <- ifelse(is.null(from), "NULL", from)
        to_str <- ifelse(is.null(to), "NULL", to)
        message("  - plotting genomic range: from=", from_str, ", to=", to_str)
    }

    ## Genome axis.
    tracks <- list(Gviz::GenomeAxisTrack())

    ## Transcript tracks (we create 1 track per transcript).
    track_names <- mcols(x)$tx_id
    if (is.null(track_names))
        track_names <- names(x)
    tx_tracks <- lapply(seq_along(x),
                        function(i) {
                          tx <- x[[i]]
                          Gviz::AnnotationTrack(tx, name=track_names[i],
                                                fill="orange", shape="box")
                        })
    tracks <- c(tracks, tx_tracks)

    ## Tracks of compatible and incompatible reads.
    if (!is.null(reads)) {
        if (!(is(reads, "GAlignments") || is(reads, "GAlignmentPairs")))
            stop("'reads' must be a GAlignments or GAlignmentPairs object")
        if (length(reads) != 0L && !(is.null(from) && is.null(to))) {
            if (is.null(from)) {
                start(plot_range) <- min(start(reads))
            } else {
                start(plot_range) <- from
            }
            if (is.null(to)) {
                end(plot_range) <- max(end(reads))
            } else {
                end(plot_range) <- to
            }
            reads <- .subsetByOverlapWithSingleRange(reads, plot_range)
        }
        reads_len <- length(reads)
        message("  - nb of reads to plot (i.e. overlapping with that ",
                "range): ", reads_len)
        if (reads_len > max.plot.reads) {
            message("  - plotting only ", max.plot.reads, " randomly chosen",
                    " reads (use the 'max.plot.reads' argument\n",
                    "    to change that limit, and/or use the 'from' and 'to'",
                    " arguments to narrow\n    down the region to plot)")
            #idx <- seq_len(max.plot.reads)
            set.seed(11L)
            idx <- order(sample(reads_len, max.plot.reads))
            reads <- reads[idx]
        }
        if (is(reads, "GAlignments")) {
            grl <- grglist(reads, order.as.in.query=TRUE)
        } else {
            grl <- grglist(reads)
        }
        ov0 <- findOverlaps(grl, x, ignore.strand=TRUE)
        ovenc0 <- encodeOverlaps(grl, x, hits=ov0,
                                 flip.query.if.wrong.strand=TRUE)
        ov0_is_compat <- isCompatibleWithSplicing(ovenc0)
        ov <- ov0[ov0_is_compat]
        is_compat_read <- countQueryHits(ov) != 0L

        ## Set strand to *.
        strand(grl@unlistData) <- "*"

        ## Set group id (required by the Gviz package, why?)
        mcols(grl@unlistData)$group <- rep.int(seq_along(grl), 
                                               elementNROWS(grl))

        ## Track of compatible reads.
        compat_grl <- grl[is_compat_read]
        name <- ifelse(length(compat_grl) == 1L,
                       names(compat_grl)[1L], "compatible reads")
        compat_reads_track <- Gviz::AnnotationTrack(compat_grl, name=name,
                                                    fill="blue", shape="box")
        tracks <- c(tracks, list(compat_reads_track))

        ## Track of incompatible reads.
        incompat_grl <- grl[!is_compat_read]
        name <- ifelse(length(incompat_grl) == 1L,
                       names(incompat_grl)[1L], "incompatible reads")
        incompat_reads_track <- Gviz::AnnotationTrack(incompat_grl, name=name,
                                                      fill="lightblue",
                                                      shape="box")
        tracks <- c(tracks, list(incompat_reads_track))
    }

    Gviz::plotTracks(tracks, from=from, to=to)
}

setGeneric("plotTranscripts", signature="x",
    function(x, reads=NULL, from=NA, to=NA, max.plot.reads=200)
        standardGeneric("plotTranscripts")
)

setMethod("plotTranscripts", "GRangesList", .plotTranscripts.GRangesList)

setMethod("plotTranscripts", "TxDb",
    function(x, reads=NULL, from=NA, to=NA, max.plot.reads=200)
    {
        ex_by_tx <- exonsBy(x, by="tx", use.names=TRUE)
        plotTranscripts(ex_by_tx, reads=reads,
                        from=from, to=to, max.plot.reads=max.plot.reads)
    }
)

setMethod("plotTranscripts", "SplicingGraphs",
    function(x, reads=NULL, from=NA, to=NA, max.plot.reads=200)
    {
        if (length(x) != 1L)
            stop("'x' must be a SplicingGraphs object of length 1")
        plotTranscripts(x[[1L]], reads=reads,
                        from=from, to=to, max.plot.reads=max.plot.reads)
    }
)