pybbi

Python interface to Jim Kent's Big Binary Indexed file (BBI) [1] library from the UCSC Genome Browser source tree using Cython.

This provides read-level access to local and remote bigWig and bigBed files but no write capabilitites. The main feature is fast retrieval of range queries into numpy arrays.

Installation

Wheels for pybbi are available on PyPI for Python 3.8, 3.9, 3.10, 3.11 on Linux (x86_64 and aarch64) and Mac OSX (x86_64/Intel). Apple Silicon (arm64) wheels will be made available once M1 runners are available in GitHub Actions.

$ pip install pybbi

API

The bbi.open function returns a BBIFile object.

bbi.open(path) -> BBIFile

path can be a local file path (bigWig or bigBed) or a URL. BBIFile objects are context managers and can be used in a with statement to clean up resources without calling BBIFile.close().

>>> with bbi.open('bigWigExample.bw') as f:
...     x = f.fetch('chr21', 1000000, 2000000, bins=40)

Introspection

BBIFile.is_bigwig -> bool
BBIFile.is_bigbed -> bool
BBIFile.chromsizes -> OrderedDict
BBIFile.zooms -> list
BBIFile.info -> dict
BBIFile.schema -> dict
BBIFile.read_autosql() -> str

Note: BBIFile.schema['dtypes'] provides numpy data types for the fields in a bigWig or bigBed (matched from the autoSql definition).

Interval output

The actual interval records in a bigWig or bigBed can be retrieved as a pandas dataframe or as an iterator over records as tuples. The pandas output is parsed according to the file's schema.

BBIFile.fetch_intervals(chrom, start, end) -> pandas.DataFrame
BBIFile.fetch_intervals(chrom, start, end, iterator=True) -> interval iterator

Summary bin records at each zoom level are also accessible.

BBIFile.fetch_summaries(chrom, start, end, zoom) -> pandas.DataFrame

Array output

Retrieve quantitative signal as an array. The signal of a bigWig file is obtained from its "value" field. The signal of a bigBed file is obtained from the genomic coverage of its intervals.

For a single range query:

BBIFile.fetch(chrom, start, end, [bins [, missing [, oob, [, summary]]]]) -> 1D numpy array

To produce a stacked heatmap from a list of (1) equal-length intervals or (2) arbitrary-length intervals with bins specified:

BBIFile.stackup(chroms, starts, ends, [bins [, missing [, oob, [, summary]]]]) -> 2D numpy array

• Summary querying is supported by specifying the number of bins for coarsening. The summary statistic can be one of: 'mean', 'min', 'max', 'cov', 'std', 'or 'sum'. (default = 'mean'). Intervals need not have the same length, in which case the data from each interval will be interpolated to the same number of bins (e.g., gene bodies).

• Missing data can be filled with a custom fill value, missing (default = 0).

• Out-of-bounds ranges (i.e. start less than zero or end greater than the chromosome length) are permitted because of their utility e.g., for generating vertical heatmap stacks centered at specific genomic features. A separate custom fill value, oob can be provided for out-of-bounds positions (default = NaN).

Function API

The original function-based API is still available:

bbi.is_bbi(path: str) -> bool
bbi.is_bigwig(path: str) -> bool
bbi.is_bigbed(path:str) -> bool
bbi.chromsizes(path: str) -> OrderedDict
bbi.zooms(path: str) -> list
bbi.info(path: str) -> dict
bbi.fetch_intervals(path: str, chrom: str, start: int, end: int, iterator: bool) -> Union[Iterable, pd.DataFrame]
bbi.fetch(path: str, chrom: str, start: int, end: int, [bins: int [, missing: float [, oob: float, [, summary: str]]]]) -> np.array[1, 'float64']
bbi.stackup(path: str, chroms: np.array, starts: np.array, ends: np.array, [bins: int [, missing: float [, oob: float, [, summary: str]]]]) -> np.array[2, 'float64']

See the docstrings for complete documentation.

Related projects

• libBigWig: Alternative C library for bigWig and bigBed files by Devon Ryan
• pyBigWig: Python bindings for libBigWig by the same author
• bw-python: Alternative Python wrapper to libBigWig by Brent Pederson
• bx-python: Python bioinformatics library from James Taylor's group that includes tools for bbi files.

This library provides bindings to the reference UCSC bbi library code. Check out @dpryan79's libBigWig for an alternative and dedicated C library for big binary files. pyBigWig also provides numpy-based retrieval and bigBed support.

References

[1]: http://bioinformatics.oxfordjournals.org/content/26/17/2204.full

From source

If wheels for your platform or Python version aren't available or you want to develop, you'll need to install pybbi from source. The source distribution on PyPI ships with (slightly modified) kent utils source, which will compile before the extension module is built.

Requires

• Platform: Linux or Darwin (Windows Subsystem for Linux seems to work too)
• pthreads, zlib, libpng, openssl, make, pkg-config
• Python 3.6+
• numpy and cython

For example, on a fresh Ubuntu instance, you'll need build-essential, make, pkg-config, zlib1g-dev, libssl-dev, libpng16-dev.

On a Centos/RedHat (rpm) system you'll need gcc, make, pkg-config, zlib-devel, openssl-devel, libpng-devel.

On a Mac, you'll need Xcode and to brew install pkg-config openssl libpng.

For development, clone the repo and install in editable mode:

$ git clone https://github.com/nvictus/pybbi.git
$ cd pybbi
$ pip install -e .

You can use the ARCH environment variable to specify a target architecture or ARCHFLAGS on a Mac.

Notes

Unfortunately, Kent's C source is not well-behaved library code, as it is littered with error calls that call exit(). pybbi will catch and pre-empt common input errors, but if somehow an internal error does get raised, it will terminate your interpreter instance.