gos 🦆

gos is a declarative (epi)genomics visualization library for Python. It is built on top of the Gosling JSON specification, providing an simplified interface for authoring interactive genomic visualizations.

Installation

Here be dragons 🐉

The API is experimental and under active development. Feedback is much appreciated and most welcomed.

pip install gosling

Example

import gosling as gos
from gosling.data import multivec

data = multivec(
    url="https://server.gosling-lang.org/api/v1/tileset_info/?d=cistrome-multivec",
    row="sample",
    column="position",
    value="peak",
    categories=["sample 1", "sample 2", "sample 3", "sample 4"],
    binSize=5,
)

base_track = gos.Track(data, width=800, height=100)

heatmap = base_track.mark_rect().encode(
    x=gos.Channel("start:G", axis="top"),
    xe="end:G",
    row=gos.Channel("sample:N", legend=True),
    color=gos.Channel("peak:Q", legend=True),
)

bars = base_track.mark_bar().encode(
    x=gos.Channel("position:G", axis="top"),
    y="peak:Q",
    row="sample:N",
    color=gos.Channel("sample:N", legend=True),
)

lines = base_track.mark_line().encode(
    x=gos.Channel("position:G", axis="top"),
    y="peak:Q",
    row="sample:N",
    color=gos.Channel("sample:N", legend=True),
)

gos.vertical(heatmap, bars, lines).properties(
    title="Visual Encoding",
    subtitle="Gosling provides diverse visual encoding methods",
    layout="linear",
    centerRadius=0.8,
    xDomain=gos.Domain(chromosome="1", interval=[1, 3000500]),
)

Local data

Data sources for the Gosling specification are expected to be accessible via HTTP. Loading a local dataset can be challenging since it requires starting a web-server and/or a Higlass server for some pre-aggregated datasets. gos provides an experimental module that transparently serves data via a background ASGI server. The various data utilites are imported from the gosling.experimental.data module.

import gosling as gos
from gosling.experimental.data import bam, csv, bigwig # file resources
from gosling.experimental.data import beddb, vector, matrix, multivec # higlass tile resources

In order to use these utilities, you will need to install additional dependencies via:

pip install "gosling[all]"
pip install clodius # optional, required for higlass tile resources

In the example above, we can replace the remote Higlass server URL with a local path to the corresponding cistrome multivec file (https://s3.amazonaws.com/gosling-lang.org/data/cistrome.multires.mv5, 4GB). gos automatically detects the local file and will starts a background Higlass server to power the visualization.

import gosling as gos
from gosling.experimental.data import multivec

data = multivec(
-   url="https://server.gosling-lang.org/api/v1/tileset_info/?d=cistrome-multivec",
+   url='../data/cistrome.multires.mv5', # path to local multivec
    row="sample",
    column="position",
    value="peak",
    categories=["sample 1", "sample 2", "sample 3", "sample 4"],
    binSize=4,
)

base_track = gos.Track(data, width=800, height=100)

Note that the visualizations will only render as long as your Python session is active.

Example Gallery

We have started a gallery of community examples in example/. If you are intereseted in contributing, please feel free to submit a PR! Checkout the existing JSON examples if you are looking for inspiration.

Development

pip install -e .

The schema bindings (gosling/schema/) and JS static assets (gosling/static/) are automatically generated using the following scripts. Please do not edit these files directly.

# generate gosling/schema/*
python tools/generate_schema_wrapper.py

# generate gosling/static/{widget.js,index.js} from src/{widget.ts,index.ts}
yarn install && yarn build:js

# Only run this if editing/using gos.GoslingWidget
jupyter nbextension install --py --symlink --overwrite --sys-prefix gosling
jupyter nbextension enable gosling --py --sys-prefix

Design & implemenation

gos is inspired by and borrows heavily from Altair both in project philosophy and implementation. The internal Python API is auto-generated from the Gosling specification using code adapted directly from Altair to generate Vega-Lite bindings. This design choice gaurentees that visualizations are type-checked in complete concordance with the Gosling specification, and that the Python API remains consitent with the evolving schema over time. Special thanks to Jake Vanderplas and others on schemapi.