Coverage and interval pileup

Show how overlapping intervals produce a coverage profile.

The top track counts how many intervals cover each position. The bottom track places overlapping intervals on separate rows. Zooming either track keeps their horizontal positions aligned.

Data use and provenance

This adapts the upstream synthetic example with a deterministic mix of interval lengths instead of random lengths, so it looks the same on each load. NumPy creates the interval starts. GenomeSpy calculates their ends, coverage, and row layout in the browser.

Adapted from the official GenomeSpy example.

Code

"""Coverage and interval pileup.

Show how overlapping intervals produce a coverage profile.
"""

import numpy as np
import pandas as pd

import genome_spy as gs


data = pd.DataFrame({"start": np.arange(1, 100)})

# Count overlaps above and arrange the same intervals into rows below.
coverage = (
    gs.Chart()
    .transform_coverage(start="start", end="end", as_="coverage")
    .mark_rect()
    .encode(x=gs.X("start:I"), x2=gs.X2("end"), y=gs.Y("coverage:Q"))
)
pileup = (
    gs.Chart()
    .transform_pileup(start="start", end="end", as_="lane")
    .mark_rect()
    .encode(
        x=gs.X("start:I"),
        x2=gs.X2("end"),
        y=gs.Y("lane:I").scale(padding=0.2, reverse=True, zoom=False),
    )
)
chart = (
    (coverage & pileup)
    .properties(data=data)
    # A deterministic mix of lengths keeps the example repeatable.
    .transform_calculate(end=gs.datum.start + 1 + (gs.datum.start * 7) % 20)
    .resolve_scale(x="shared")
)

Show generated specification