Sashimi plot

A splice-junction view with coverage in the background and arc links connecting junction endpoints. Labels on the arcs show junction support while the signal track keeps local expression context visible.

Data use and provenance

The example is based on igv.js’s splice-junction sample data for chr15. The source files are hosted in IGV’s data repository, and the original demo is available here: splice junction example.

What to notice

Arc height follows junction span, while stroke thickness and label text encode uniquely mapped reads. Horizontal zoom changes the shared arc-and-label height domain; the coverage track retains an independent y scale.

Python implementation

The chart layers lazy BigWig coverage with filtered junction links and labels. Formula transforms derive arc geometry and deterministic label jitter, and an expression-driven scale couples the junction height to the visible locus.

See the official GenomeSpy example for the source visualization and scale-expression details.

Code

"""Sashimi plot.

A splice-junction view with coverage in the background and arc links connecting
junction endpoints. Labels on the arcs show junction support while the signal
track keeps local expression context visible.
"""

from __future__ import annotations

import genome_spy as gs

# Add a slider to hide junctions supported by fewer reads.
min_uniquely_mapped_reads = gs.param(
    "minUniquelyMappedReads",
    value=1,
    bind=gs.binding_range(
        min=0,
        max=200,
        step=1,
        name="Min uniquely mapped reads",
    ),
)

DOMAIN = [
    {"chrom": "chr15", "pos": 92925000},
    {"chrom": "chr15", "pos": 92949000},
]

# Show read coverage in grey, loading only the region being viewed.
coverage = (
    gs.Chart(
        gs.lazy.bigwig(
            "https://raw.githubusercontent.com/igvteam/igv-data/refs/heads/main/data/test/splice_junctions/splice_junction_track_test_cases_sampleA.chr15-92835700-93031800.bigWig",
            pixelsPerBin=1,
        )
    )
    .mark_rect(color="lightgray", minWidth=0.5, minOpacity=1, tooltip=None)
    .transform_filter(gs.datum.score > 0)
    .encode(
        x=gs.Locus("chrom", "start"),
        x2=gs.Locus("chrom", "end"),
        y=gs.Y("score:Q").scale(nice=True, zero=False, reverse=False).title("Coverage"),
    )
    .properties(name="coverage")
)

# Connect splice junctions with arcs: longer spans rise higher, and more reads
# make the lines thicker.
arc_layer = (
    gs.Chart()
    .mark_link(linkShape="dome", maxChordLength=100000000)
    .encode(
        x=gs.Locus("chrom", "chromStart", band=0),
        x2=gs.Locus("chrom", "chromEnd", band=0),
        y=gs.Y("span:Q")
        .scale(
            type="sqrt",
            domain=gs.expr("[0, span(domain('x')) * height / width * 5]"),
            reverse=False,
        )
        .axis(None),
        size=gs.Size("score:Q").scale(type="sqrt", range=[0.1, 2.0]),
    )
    .properties(name="arcs")
)

# Label each arc with its supporting read count.
label_layer = (
    gs.Chart()
    .mark_text(dy=-8, tooltip=False)
    .transform_formula(
        expr=(gs.datum.chromEnd + gs.datum.chromStart) / 2,
        as_="center",
    )
    .encode(
        x=gs.Locus("chrom", "center"),
        y=gs.Y("span:Q").scale(reverse=False),
        text=gs.Text("score:Q"),
    )
    .properties(name="labels")
)

# Load junction positions and apply the slider's minimum read count.
splice_junctions = (
    gs.layer(arc_layer, label_layer)
    .properties(
        name="splice-junctions",
        data=gs.Data(
            url="https://raw.githubusercontent.com/igvteam/igv-data/refs/heads/main/data/test/splice_junctions/splice_junction_track_test_cases_sampleA.chr15-92835700-93031800.SJ.out.bed",
            format=gs.data_format(type="bed"),
        ),
    )
    .transform_filter(gs.datum.score >= min_uniquely_mapped_reads)
    .transform_formula(expr=gs.datum.chromEnd - gs.datum.chromStart, as_="span")
    # Vary arc heights slightly to help separate nearby junctions.
    .transform_formula(
        expr=gs.datum.span + (gs.datum.span % 10 - 5) / 10 * gs.datum.span,
        as_="span",
    )
)

# Draw the arcs over the coverage and attach the slider.
chart = (
    gs.layer(coverage, splice_junctions)
    .properties(
        assembly="hg38",
        title="Sashimi plot",
        description=(
            "A sashimi-style splice junction view adapted from the GenomeSpy "
            "docs, using lazy BigWig coverage and splice-junction arcs."
        ),
        scales=gs.scales(x=gs.Scale(domain=DOMAIN)),
    )
    .add_params(min_uniquely_mapped_reads)
    .resolve_scale(y="independent")
    .resolve_axis(y="independent")
)

Show specification