/sambamba

Tools for working with SAM/BAM data

Primary LanguageDGNU General Public License v2.0GPL-2.0

Build Status AnacondaBadge DL BrewBadge GuixBadge DebianBadge

A. Tarasov, A. J. Vilella, E. Cuppen, I. J. Nijman, and P. Prins. Sambamba: fast processing of NGS alignment formats. Bioinformatics, 2015.

SAMBAMBA

Table of Contents

Introduction

Sambamba is a high performance highly parallel robust and fast tool (and library), written in the D programming language, for working with SAM and BAM files. Because of its efficiency Sambamba is an important work horse running in many sequencing centres around the world today. As of December 2021, Sambamba has been cited over 740 times and has been installed from Conda over 200K times. Sambamba is also distributed by Debian. To cite sambamba see Credit.

Current functionality is an important subset of samtools functionality, including view, index, sort, markdup, and depth. Most tools support piping: just specify /dev/stdin or /dev/stdout as filenames. When we started writing sambamba (in 2012) the main advantage over samtools was parallelized BAM reading and writing.

In March 2017 samtools 1.4 was released, reaching some parity. A recent performance comparison shows that sambamba still holds its ground and can even do better. Here are some comparison metrics. For example, for flagstat sambamba is 1.4x faster than samtools. For index they are similar. For markdup almost 6x faster and for view 4x faster. For sort sambamba has been beaten, though sambamba is notably up to 2x faster than samtools on large RAM machines (120GB+).

In addition sambamba has a few interesting features to offer, in particular

  • fast large machine sort, see performance
  • automatic index creation when writing any coordinate-sorted file
  • view -L <bed file> utilizes BAM index to skip unrelated chunks
  • depth allows to measure base, sliding window, or region coverages
    • Chanjo builds upon this and gets you to exon/gene levels of abstraction
  • markdup, a fast implementation of Picard algorithm
  • slice quickly extracts a region into a new file, tweaking only first/last chunks
  • and more (you'll have to try)

The D language is extremely suitable for high performance computing (HPC). At this point we think that the BAM format is here to stay for processing reference guided sequencing data and we aim to make it easy to parse and process BAM files.

Sambamba is free and open source software, licensed under GPLv2+. See manual pages online to know more about what is available and how to use it.

For more information on Sambamba contact the mailing list (see Getting help).

No CRAM support

Important notice: with version 0.8 support for CRAM was removed from Sambamba (see the RELEASE NOTES)

To use CRAM you can still use one of the older (binary) releases of Sambamba.

Binary installation

Install stable release

For those not in the mood to learn/install new package managers, there are Github source and binary releases. Simply download the tarball, unpack it and run it according to the accompanying release notes.

Below package managers Conda, GNU Guix, Debian and Homebrew also provide recent binary installs for Linux. For MacOS you may use Conda or Homebrew.

Bioconda install

Install with CONDA

Ther should be binary downloads for Linux and MacOS.

With Conda use the bioconda channel.

GNU Guix install

GuixBadge

A reproducible GNU Guix package for sambamba is available. The development version is packaged here.

Debian GNU/Linux install

DebianBadge

See also Debian package status.

Homebrew install

BrewBadge

Users of Homebrew can also use the formula from homebrew-bio.

brew install brewsci/bio/sambamba

It should work for Linux and MacOS.

Getting help

Sambamba has a mailing list for installation help and general discussion.

Reporting a sambamba bug or issue

Before posting an issue search the issue tracker and mailing list first. It is likely someone may have encountered something similar. Also try running the latest version of sambamba to make sure it has not been fixed already. Support/installation questions should be aimed at the mailing list. The issue tracker is for development issues around the software itself. When reporting an issue include the output of the program and the contents of the output directory.

Please use the following check list. It exists for multiple reasons :)

Check list:

  1. I have found and issue with sambamba
  2. I have searched for it on the issue tracker (also check closed issues)
  3. I have searched for it on the mailing list
  4. I have tried the latest release of sambamba
  5. I have read and agreed to below code of conduct
  6. If it is a support/install question I have posted it to the mailing list
  7. If it is software development related I have posted a new issue on the issue tracker or added to an existing one
  8. In the message I have included the output of my sambamba run
  9. In the message I have included the relevant files in the output directory
  10. I have made available the data to reproduce the problem (optional)

To find bugs the sambamba software developers may ask to install a development version of the software. They may also ask you for your data and will treat it confidentially. Please always remember that sambamba is written and maintained by volunteers with good intentions. Our time is valuable too. By helping us as much as possible we can provide this tool for everyone to use.

Code of conduct

By using sambamba and communicating with its communtity you implicitely agree to abide by the code of conduct as published by the Software Carpentry initiative.

Compiling Sambamba

Note: in general there is no need to compile sambamba. You can use a recent binary install as listed above.

The preferred method for compiling Sambamba is with the LDC compiler which targets LLVM. LLVM version 6 is faster than earlier editions.

Compilation dependencies

See INSTALL.md.

Compiling for Linux

The LDC compiler's github repository provides binary images. The current preferred release for sambamba is LDC - the LLVM D compiler (>= 1.6.1). After installing LDC from https://github.com/ldc-developers/ldc/releases/ with, for example

cd
wget https://github.com/ldc-developers/ldc/releases/download/v$ver/ldc2-1.7.0-linux-x86_64.tar.xz
tar xvJf ldc2-1.7.0-linux-x86_64.tar.xz
export PATH=$HOME/ldc2-1.7.0-linux-x86_64/bin:$PATH
export LIBRARY_PATH=$HOME/ldc2-1.7.0-linux-x86_64/lib
git clone --recursive https://github.com/biod/sambamba.git
cd sambamba
make

To build a development/debug version run

make clean && make debug

To run the test fetch shunit2 from https://github.com/kward/shunit2 and put it in the path so you can run

make check

See also INSTALL.md.

GNU Guix

Our development and release environment is GNU Guix. To build sambamba the LDC compiler is also available in GNU Guix:

guix package -A ldc

For more instructions see INSTALL.md.

Compiling with Meson/Ninja

Debian uses a meson+ninja build. It may work with something like

meson build
cd build
ninja
ninja test

or with some tuning

rm -rf build/ ; env D_LD=gold CC=gcc meson build --buildtype release
cd build/
env CC=gcc ninja
env CC=gcc ninja test

Compiling for MacOS

Sambamba builds on MacOS. We have a Travis integration test as an example. It can be something like

    brew install ldc
    git clone --recursive https://github.com/biod/sambamba.git
    cd sambamba
    make

Development

Sambamba development and issue tracker is on github. Developer documentation can be found in the source code and the development documentation.

Debugging and troubleshooting

Segfaults on certain Intel Xeons

Important note: some older Xeon processors segfault under heavy hyper threading - which Sambamba utilizes. Please read this when encountering seemingly random crashes. There is no real fix other than disabling hyperthreading. Also discussed here. Thank Intel for producing this bug.

Dump core

In a crash sambamba can dump a core file. To make this happen set

ulimit -c unlimited

and run your command. Send us the core file so we can reproduce the state at time of segfault.

Use catchsegv

Another option is to use catchsegv

catchsegv ./build/sambamba command

this will show state on stdout which can be sent to us.

Using gdb

In case of crashes it's helpful to have GDB stacktraces (bt command). A full stacktrace for all threads:

thread apply all backtrace full

Note that GDB should be made aware of D garbage collector which emits SIGUSR signals and gdb needs to ignore them with

handle SIGUSR1 SIGUSR2 nostop noprint

A binary relocatable install of sambamba with debug information and all dependencies can be fetched from the binary link above. Unpack the tarball and run the contained install.sh script with TARGET

./install.sh ~/sambamba-test

Run sambamba in gdb with

gdb -ex 'handle SIGUSR1 SIGUSR2 nostop noprint' \
  --args ~/sambamba-test/sambamba-*/bin/sambamba view --throw-error

License

Sambamba is generously distributed under GNU Public License v2+.

Credit

Citations are the bread and butter of Science. If you are using Sambamba in your research and want to support our future work on Sambamba, please cite the following publication: A. Tarasov, A. J. Vilella, E. Cuppen, I. J. Nijman, and P. Prins. Sambamba: fast processing of NGS alignment formats. Bioinformatics, 2015.

Bibtex reference

@article{doi:10.1093/bioinformatics/btv098,
  author = {Tarasov, Artem and Vilella, Albert J. and Cuppen, Edwin and Nijman, Isaac J. and Prins, Pjotr},
  title = {Sambamba: fast processing of NGS alignment formats},
  journal = {Bioinformatics},
  volume = {31},
  number = {12},
  pages = {2032-2034},
  year = {2015},
  doi = {10.1093/bioinformatics/btv098},
  URL = { + http://dx.doi.org/10.1093/bioinformatics/btv098}