QScores-Archiver is a system for archiving quality scores in FASTQ formatted next generation sequencing data. To achieve this, it employs lossy and lossless transformations coupled with compression. It is implemented as a command-line tool, but has been developed so that users can link this software with their own source code. This document accompanies the archive, which also includes:
- source code in C++,
- other important documentation and license information, and
- a very small data file for testing.
This archive does not include any binaries. This program is described in the paper:
R. Wan, V. N. Anh, and K. Asai. "Transformations for the Compression of FASTQ Quality Scores of Next Generation Sequencing Data". Bioinformatics, 28(5):623-635, 2012
which we refer to as "the paper" throughout this document.
The software also includes implementations of static codes, Huffman coding, and interpolative coding as separate C++ classes. See these papers or books for further details:
- I. H. Witten, A. Moffat, and T. C. Bell. "Managing Gigabytes", 1999, Second. edition, Morgan Kaufmann.
- A. Moffat and L. Stuiver. "Binary Interpolative Coding for Effective Index Compression". Information Retrieval 3(1): 25-47 (2000).
- A. Moffat and A. Turpin, "Compression and Coding Algorithms", 2002, Kluwer Academic Publishers.
The software has been updated in 2020 for current compilers.
| Software | Minimum version | Tested version | Required? | Web site |
|---|---|---|---|---|
| g++ | 5.2.1 | 13.2.0 | Yes | http://gcc.gnu.org/ |
| CMake | 3.5 | 3.28.3 | Yes | http://www.cmake.org/ |
| Boost library | 1.59.0 | 1.85.0 | Yes | http://www.boost.org/ |
| zlib library | 1.2.3 | 1.2.11 | No | http://www.zlib.net/ |
| gzip | 1.3.5 | 1.10 | No | http://www.gzip.org/ |
| libbzip library | 1.0.0 | 1.0.8 | No | http://www.bzip.org/ |
| bzip2 | 1.0.3 | 1.0.8 | No | http://www.bzip.org/ |
| Doxygen | 1.9.4 | 1.9.8 | No | http://www.stack.nl/~dimitri/doxygen/ |
Experiments in the paper using QScores-Archiver was executed on Linux systems running Debian 6.0 (squeeze) or CentOS 5.4. Currently, it is being maintained on an Ubuntu 24.04 system (i.e., it's been tested recently on such a system).
Both optional and required tools for compiling or using QScores-Archiver is listed in the table above. The column "Minimum version" refers to the software versions used during software development and when running the experiments in the paper. They do not represent the minimum requirements; it is possible that lower versions can be used. The column "Tested version" refers to the versions used for the most recent tests on Ubuntu 24.04.
The compression libraries and executables zlib, gzip, libbzip, and bzip2 are all optional and the software will compile without them.
Doxygen is a documentation system to extract comments that have been placed inline in the source code. See the section below entitled "Software Documentation" for more information.
The Boost Library must be both installed and compiled to make use of the program_options, system, and filesystem libraries. Under some Linux distributions, Boost can be installed using its associated package manager (such as apt for Debian and Ubuntu). For example, on a more recent Ubuntu 24.04 system, the following packages (and its dependencies) were installed using apt:
- libboost-program-options1.71-dev
- libboost-system1.71-dev
- libboost-filesystem1.71-dev
- libboost-serialization1.71-dev
- libboost-mpi1.71-dev
Consult the Boost documentation for further information.
A conda environment can be created to include some of the above dependencies. However, Boost (the most important dependency) is excluded at the moment.
To create the environment, type:
`conda env create -f environment.yml`
The QScores-Archiver software is written in C++ and has been compiled using versions 5.2.1 and 9.3.0 of g++. The system has been tested on a 64-bit system, but it should work on other architectures.
CMake (at least version 3.5) is used to compile the software and it is recommended that an "out-of-source" build is performed so as not to clutter the original source directories. We give some brief instructions below on how to do this:
-
Install Boost; consult the Boost documentation on up-to-date information on how to do this. Then set the variable BOOST_ROOT to the location of Boost if it has not already been set:
export BOOST_ROOT=/usr/local/boost_1_59_0/ -
Create a directory for the repository [i.e.,
~/tmp/] and clone it into there. -
Enter the
~/tmp/QScores-Archiver/directory (or whichever name you chose) and create abuild/directory. Then, enter it. (Actually,build/can be anywhere, including an entirely separate directory tree. You can delete it after compilation or you may want to keep it if you plan to re-compile the source code.). Then run,cmake ../srcwhere ".." represents the location of the top-level CMakeLists.txt in
~/tmp/QScores-Archiver/src/. By default, this will set up a Makefile to install the program into/usr/local/, which would require system administrator access. To use another directory, type this (for example):cmake .. -DCMAKE_INSTALL_PREFIX=~/tmpreplacing the installation prefix with whatever you prefer.
-
Type
maketo compile the C++ source code of QScores-Archiver. If this succeeds, then the executable should be in the build subdirectory asqscores/qscores-archiver. -
Type
make testto run through a series of tests. There are 62 tests in total and each one should say Passed. -
Finally, type
make installto install the software. This copies the important files from the archive to the installation prefix specified in the cmake line above (see "Files_and_Directories" for information about the structure) . The~/tmp/QScores-Archiver/directory, including thebuild/directory, can now be deleted, unless you are interested in viewing the source code.
The dependencies between the various modules is depicted in the figure below (generated using cmake with the --graphviz option):
QScores-Archiver was developed with inline comments that can be extracted using the Doxygen documentation system. They can be created using make as well. Here, we show how to generate the documentation for the QScores-Archiver main code.
- In the directory where you ran make to compile the executable, type
make qscores-archiver-doc. If Doxygen can be found, then the documentation will be generated in the doc/ subdirectory. - You may choose to do a
make installwhich would put this documentation in thedoc/subdirectory under the installation prefix. See "Files and Directories" for more information.
The following HTML file can be used if the documentation has been installed: documentation.html. Of course, the documentation has to be created first according to step 1.
After installation, the following directory structure should result:
. Directory you specified when you ran `cmake`
├── bin Binary files, including the qscores-archiver executable
├── data Data directory
│ ├── sample.qs Sample data file of 1,000 quality scores of length 100 qscores each
└── data Documentation directory; Doxygen-generated documents are placed here
├── classes.png Directed graph of dependencies between libraries.
├── classes.dia Graphviz source code for classes.png .
├── documentation.html Top-level HTML file for viewing Doxygen-generated documents
├── COPYING Copy of GNU GPL license v3
├── COPYING.DOC Copy of GNU FDL license v1.3
├── COPYING.LESSER Copy of GNU LGPL license v3
├── AUTHORS Contributors to the software
└── ChangeLog History of changes
If no installation prefix was provided and the installation is performed with root access, then system defaults are used instead.
[*] Source code is not copied when you do a "make install". If you are interested in examining the source, then look under the src/ directory in the expanded archive.
Three lossy transformations have been incorporated into QScores-Archiver. To facilitate the discussion in the paper, their parameters have been unified as |Σ| -- the size of the new alphabet after the transformation. In the program, they actually each have a different parameter, as outlined in the next table.
| Lossy method | Abbreviation | Parameter | Meaning |
|---|---|---|---|
| Truncation | trunc | --trunc x | All qscores larger than x are changed_to_x. |
| Uniform binning | unibinning | --unibin x | Bins the error probabilities in a uniform manner by dividing the range of probabilities [0, 1] into_x_bins. |
| Logarithmic binning | logbinning | --logbin x | Bins the quality scores (which themselves are on a logarithmic scale) so that each bin has x values at most. Binning starts from the lowest PHRED score (i.e., highest error probability). |
The potential confusion is that x has a different meaning in each case but end up achieving the same thing -- reducing the size of the quality scores' alphabet. This reduced alphabet is used for the graphs in the paper. A one-to-many mapping exists between x for each lossy method and |Σ|. For truncation, the mapping is obvious. For the other two cases, in the doc/ directory, we provide two text files:
- unibinning.txt
- logbinning.txt
Both of these files are two-column tables. The first column is the parameter to give to QScores-Archiver (i.e., x) and the second column is the size of the alphabet (i.e., |Σ|). Note that ties in the alphabet size are possible -- logarithmic binning with either 12 quality scores per bin or 13 quality scores per bin will results in an alphabet of size 8. For the paper, even duplicate |Σ| were plotted -- the difference is just too small to notice in the graphs.
A sample data file "data/sample.qs" is included in the data/ subdirectory. It is a list of 1,000 randomly generated "quality scores" of length 100 each (since they are randomly generated, it is more correct to call them a sequence of characters whose range is stipulated by the Sanger-FASTQ range). Assuming the software has been installed, then this data file and the executable are in the same parent directory. Here are some example executions:
-
Display a list of available options.
./qscores-archiver --help
-
Show the status of externally installed software (i.e., zlib, gzip, etc.)
./qscores-archiver --checkextern
-
Examine the bins created from using logarithmic binning with 2 qscores per bin and Sanger-FASTQ format.
./qscores-archiver --checkbin --mapping sanger --logbin 2
-
Examine the reversed bins created from uniform binning with the interval [0, 1] divided into 10,000 bins.
./qscores-archiver --checkunbin --mapping sanger --unibin 10000
-
Binary encode the test file with verbose output.
./qscores-archiver --input ../data/sample.qs --output test.qs --encode --binary --verbose
-
Binary decode the compressed test file with verbose output
./qscores-archiver --input test.qs --output test-out.qs --decode --verbose
-
Do not perform any transformations or compressions (i.e., the data file is unchanged and is identical to the output).
./qscores-archiver --input ../data/sample.qs --output test.qs --encode --nocompress
-
Perform logarithmic binning with 10 qscores per bin and Gamma code the result. Also, create blocks of 10 reads each.
./qscores-archiver --input ../data/sample.qs --output test.qs --encode --gamma --logbin 10 --blocksize 10
If decoding is being performed and --nocompress was not used, then the transformation and compression options are included in the compressed file. So, they do not need to be provided when decompressing. Obviously, if --nocompress was selected, then the output cannot be decompressed. This option's purpose is to see the output from the lossy transformations; to make use of them, note that 1-based bin numbers have been encoded. You will need to add 32 to each value to put them into Sanger-FASTQ format.
There are many things that were intended for QScores-Archiver which have not yet been implemented. For example, additional compression methods such as Re-Pair [1,2], Arithmetic coding, and Prediction by Partial Matching were considered. They may still be implemented in the future if there is enough interest from users.
Also, QScores-Archiver does not make use of standard input and output. To be honest, I tried and did not know how in C++ for binary input/output. However, since this would be a useful feature to have to reduce disk I/O if QScores-Archiver is used in a pipeline, this remains a priority for me.
[1] N. J. Larsson and A. Moffat. Offline Dictionary-Based Compression. In Proc. IEEE, 88(11), 1722-1732, November 2000.
[2] See also [Re-Store](http://rwanwork.info/en/restore.html).
This software was implemented while I was at the Computational Biology Research Centre (Tokyo, Japan) and the University of Tokyo (around 2010-2011). My contact details:
E-mail: rwan.work@gmail.com
My homepage is here.
The latest version of QScores-Archiver can be downloaded from GitHub.
If you have any information about bugs, suggestions for the documentation or just have some general comments, feel free to contact me via e-mail or GitHub.
QScores-Archiver (Quality scores archiver)
Copyright (C) 2011-2015, 2024 by Raymond Wan
QScores-Archiver is distributed under the terms of the GNU Lesser General Public License (LGPL, version 3 or later) -- see the file COPYING and COPYING.LESSER for details.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.3 or any later version published by the Free Software Foundation; with no Invariant Sections, no Front-Cover Texts and no Back-Cover Texts. A copy of the license is included with the archive as COPYING.DOC.
Please see the doc/ directory for license and documentation.
This GitHub repository was created from the original tarball on my homepage a few years ago. Initially, it was identical to the version described in the paper. Hopefully, it will be easier for me to maintain in GitHub.
Raymond Wan
October 7, 2020