This repository contains a reference implementation of the BGEN format, written in C++. The library can be used as the basis for BGEN support in other software, or as a reference for developers writing their own implementations of the BGEN format.
This repository contains the library itself, a set of example data files, and a number of example programs (e.g. bgen_to_vcf) that demonstrate the use of the library API.
In addition, a number of utilities built using the library are also included in this repository:
- bgenix - a tool to index and efficiently retrieve subsets of a BGEN file.
- cat-bgen - a tool to efficiently concatenate BGEN files.
- edit-bgen - a tool to edit BGEN file metadata.
- An R package called rbgen is also constructed in the build directory. See the rbgen wiki page for more information on using this package.
If you make use of the BGEN library, its tools or example programs, please cite:
Band, G. and Marchini, J., "BGEN: a binary file format for imputed genotype and haplotype data", bioArxiv bioRxiv 308296; doi: https://doi.org/10.1101/308296
Thanks!
This BGEN implementation is released under the Boost Software License v1.0. This is a relatively permissive open-source license that is compatible with many other open-source licenses. See this page and the file LICENSE_1_0.txt for full details.
This repository also contains code from the sqlite, boost, and zstandard libraries, which comes with their own respective licenses. (respectively, public domain, the boost software license, and the BSD license). These libraries are not used in the core BGEN implementation, but may be used in the example programs provided.
The UK Biobank has released imputed genotype data for almost half a million individuals in BGEN format, with accompanying bgenix index files. The original release of this data (version 2) had an issue with naming of the index files. Please see [here](https://bitbucket.org/gavinband/bgen/wiki/Using the UK Biobank full release index files) for information on working around this. The more recent version of this data (version 3) does not have this issue.
The following commands (typed into a UNIX shell - the dollar symbol indicates the prompt, and shouldn't be typed in) should perform a basic download and install of the BGEN library, example data and tools:
$ # get it
$ wget http://bitbucket.org/gavinband/bgen/get/master.tar.gz
$ cd bgen
$ # compile it
$ ./waf configure
$ ./waf
$ # test it
$ ./build/test/unit/test_bgen
$ ./build/apps/bgenix -g example/example.16bits.bgen -listThe following sections contains more information on this process.
A tarball of the latest master branch is available here: http://bitbucket.org/gavinband/bgen/get/master.tar.gz.
Alternatively, use mercurial to download the master branch as follows:
hg clone https://[email protected]/gavinband/bgen -u master(This command can take a while.)
Additionally, pre-built version of the bgen utilities may be available from this page. Note: the recommended use is to download and compile bgenix for your platform; these binaries are provided for convenience in getting started quickly.
To compile the code, use the supplied waf build tool:
./waf configure
./wafResults will appear under the build/ directory.
Note: a full build requires a compiler that supports C++11, e.g. gcc v4.7 or above. To specify the compiler used, set the CXX environment variable during the configure step. For example (if your shell is bash):
CXX=/path/to/g++ ./waf configure
./waf
The sqlite and zstd libraries are written in C; to specify the C compiler you can additionally add CC=/path/to/gcc. We have tested compilation on gcc 4.9.3 and 5.4.0, and using clang, among others.
If you don't have access to a compiler with C++11 support, you can still build the core bgen implementation, but won't be able to build the applications or example programs. See the wiki for more information.
BGEN's tests can be run by typing
./build/test/test_bgenor, for more recent versions:
./build/test/unit/test_bgenIf all goes well a message like All tests passed should be printed.
If you have Robot Test Framework installed, you can instead run the full suite of unit and functional tests like so:
./test/functional/run_tests.shTest results will be placed in the directory build/test/functional/test-reports.
The example program bgen_to_vcf reads a bgen file (v1.1 or v1.2) and outputs it as a VCF file to stdout. You can try running it
by typing
./build/example/bgen_to_vcf example/example.8bits.bgenwhich should output vcf-formatted data to stdout. We've provided further example bgen files in the example/ subdirectory.
The command
./waf installwill install the applications listed above into a specified system or user directory. By default this is /usr/local. To change it, specify the prefix at the configure step:
./waf configure --prefix=/path/to/installation/directory
./waf installThe programs listed above will be installed into a folder called bin/ under the prefix dir, e.g. bgenix will be installed as /path/to/installation/directory/bin/bgenix etc.
Note that in many cases there's no need for installation; the executables are self-contained. The install step simply copies them into the destination directory.
(The installation prefix need not be a system-wide directory. For example, I typically specify an installation directory within my home dir, e.g. ~gav/projects/software/.
This repo follows the branch naming practice in which master represents the most up-to-date code considered in a 'releasable' state. If you are interested in using bgen code in your own project, we therefore recommend cloning the master branch. Code development takes place in the default branch and/or in feature branches branched from the default branch. The command given above downloads the master branch, which is what most people will want.
See the source code, BGEN releases, or the Wiki for more information.