#SSW Library: An SIMD Smith-Waterman C/C++/Python Library for Use in Genomic Applications

License: MIT

Copyright (c) 2012-2015 Boston College

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Author: Mengyao Zhao & Wan-Ping Lee

Contact: Mengyao Zhao zhangmp@bc.edu

Last revision: 26/07/2014

##Overview SSW is a fast implementation of the Smith-Waterman algorithm, which uses the Single-Instruction Multiple-Data (SIMD) instructions to parallelize the algorithm at the instruction level. SSW library provides an API that can be flexibly used by programs written in C, C++ and other languages. We also provide a software that can do protein and genome alignment directly. Current version of our implementation is ~50 times faster than an ordinary Smith-Waterman. It can return the Smith-Waterman score, alignment location and traceback path (cigar) of the optimal alignment accurately; and return the sub-optimal alignment score and location heuristically.

###How to use the API The API files include ssw.h and ssw.c, which can be directly used by any C or C++ program. For the C++ users who are more comfortable to use a C++ style interface, an additional C++ wrapper is provided with the file ssw_cpp.cpp and ssw_cpp.h.

To use the C style API, please:

1. Download ssw.h and ssw.c, and put them in the same folder of your own program files.
2. Write #include "ssw.h" into your file that will call the API functions.
3. The API files are ready to be compiled together with your own C/C++ files.

The API function descriptions are in the file ssw.h. One simple example of the API usage is example.c. The Smith-Waterman penalties need to be integers. Small penalty numbers such as: match: 2, mismatch: -1, gap open: -3, gap extension: -1 are recommended, which will lead to shorter running time.

To use the C++ style API, please:

1. Download ssw.h, ssw.c, ssw_cpp.cpp and ssw_cpp.h and put them in the same folder of your own program files.
2. Write #include "ssw_cpp.h" into your file that will call the API functions.
3. The API files are ready to be compiled together with your own C/C++ files.

The API function descriptions are in the file ssw_cpp.h. A simple example of using the C++ API is example.cpp.

###Speed and memory usage of the API Test data set: Target sequence: reference genome of E. coli strain 536 (4,938,920 nucleotides) from NCBI Query sequences: 1000 reads of Ion Torrent sequenced E. coli strain DH10B (C23-140, 318 PGM Run, 11/2011), read length: ~25-540 bp, most reads are ~200 bp

CPU time:

• AMD CPU: default penalties: ~880 seconds; -m1 -x3 -o5 -e2: ~460 seconds
• Intel CPU: default penalties: ~960 seconds; -m1 -x3 -o5 -e2: ~500 seconds

Memory usage: ~40MB

###Install the software

1. Download the software from https://github.com/mengyao/Complete-Striped-Smith-Waterman-Library.
2. cd src
3. make
4. The executable file will be ssw_test.

###Run the software

Usage: ssw_test [options] ... <target.fasta> <query.fasta>(or <query.fastq>)
Options:
	-m N	N is a positive integer for weight match in genome sequence alignment. [default: 2]
	-x N	N is a positive integer. -N will be used as weight mismatch in genome sequence alignment. [default: 2]
	-o N	N is a positive integer. -N will be used as the weight for the gap opening. [default: 3]
	-e N	N is a positive integer. -N will be used as the weight for the gap extension. [default: 1]
	-p	Do protein sequence alignment. Without this option, the ssw_test will do genome sequence alignment.
	-a FILE	FILE is either the Blosum or Pam weight matrix. [default: Blosum50]
	-c	Return the alignment path.
	-f N	N is a positive integer. Only output the alignments with the Smith-Waterman score >= N.
	-r	The best alignment will be picked between the original read alignment and the reverse complement read alignment.
	-s	Output in SAM format. [default: no header]
	-h	If -s is used, include header in SAM output.

###Software output The software can output SAM format or BLAST like format results.

1. SAM format output:

Example:

@HD VN:1.4  SO:queryname
@SQ SN:chr1 LN:1001
6:163296599:F:198;None;None/1   0   chr1    453 5   3M2D3M1D4M2D6M1D5M1D5M2I7M  *   0   0   CCAGCCCAAAATCTGTTTTAATGGTGGATTTGTGT *   AS:i:37 NM:i:11 ZS:i:28
3:153409880:F:224;None;3,153410143,G,A/1    0   chr1    523 4   2M1D32M1D3M1D6M1D8M *   0   0   GAAGAGTTAATTTAAGTCACTTCAAACAGATTACGTATCTTTTTTTTCCCT *   AS:i:42 NM:i:16 ZS:i:41
Y:26750420:R:-132;None;None/1   0   chr1    120 4   2M1I4M3D3M1I7M2I9M2D6M1I8M  *   0   0   AACAACAGAAGTTAATTAGCTTCAAAAATACTTTATATTTGCAA    *   AS:i:32 NM:i:16 ZS:i:29
13:91170622:R:-276;None;None/1  0   chr1    302 4   8M1D8M1D3M2D6M1D4M2I2M1D2M3D5M1I4M  *   0   0   CATTTATTGTTGTTTTTAAAGATTAAATGATTAAATGTTTCAAAA   *   AS:i:32 NM:i:18 ZS:i:30
15:37079528:R:-240;None;None/1  0   chr1    4   5   4M2D4M1D9M1I3M4I16M1I3M1D4M2D5M *   0   0   ACAGTGATGCCAAGCCAGTGGGTTTTAGCTTGTGGAGTTCCATAGGAGCGATGC  *   AS:i:30 NM:i:22 ZS:i:23
9:92308501:R:-176;None;None/1   0   chr1    142 4   4M3I5M4D10M2D4M1I2M2I6M5D1M1D6M2D3M *   0   0   AATAACCATAAAAATGGGCAAAGCAGCCTTCAGGGCTGCTGTTTCTA *   AS:i:26 NM:i:25 ZS:i:26
...

What is the output? Please check the document "The SAM Format Specification" at: http://samtools.sourceforge.net/SAM1.pdf for the full description. The additional optional field "ZS" indicates the suboptimal alignment score. For example, the 1st record in the upper example means the optimal alignment score of the given sequence is 37; the suboptimal alignment score is 28; the mismatch and INDEL base count within the aligned fragment of the read is 11.

2. An example of the BLAST like output:

target_name: chr1
query_name: 6:163296599:F:198;None;None/1
optimal_alignment_score: 37	sub-optimal_alignment_score: 28	strand: +	target_begin: 453	target_end: 492	query_begin: 17	query_end: 51

Target:      453    CCAATGCCACAAAACATCTGTCTCTAACTGGTG--TGTGTGT    492
                    |||  ||| ||||  |||||| | ||| |||||  |*|||||
Query:        17    CCA--GCC-CAAA--ATCTGT-TTTAA-TGGTGGATTTGTGT    51

target_name: chr1
query_name: 3:153409880:F:224;None;3,153410143,G,A/1
optimal_alignment_score: 42	sub-optimal_alignment_score: 41	strand: +	target_begin: 523	target_end: 577	query_begin: 3	query_end: 53

Target:      523    GAGAGAGAAAATTTCACTCCCTCCATAAATCTCACAGTATTCTTTTCTTTTTCCT    577
                    || ||||**|||||*|*||*||*||*|*|**|*|| ||| |||||| ||||*|||
Query:         3    GA-AGAGTTAATTTAAGTCACTTCAAACAGATTAC-GTA-TCTTTT-TTTTCCCT    53
...

##Python interface

###How to use the python wrapper ssw_wrap.py

A Python wrapper partially implements the c library functionality (only DNA sequences can be aligned for now). c libraries are completely integrated in a simple module that do not require any C programming knowledge to be used. Briefly, An aligner object can be initialized with alignment parameters and a reference subject sequence, then the object method align can be called with a query sequence and filtering conditions (min score and min length) as many time as desired. Depending of the score and length requested an python object PyAlignRes will be eventually returned.

To use the python wrapper, please:

1. Compile the scr folder by either using the makefile or by compiling a dynamic shared library with gcc gcc -Wall -O3 -pipe -fPIC -shared -rdynamic -o libssw.so ssw.c ssw.h

2. libssw.so and ssw_wrap.py can them be put in the same folder of your own program files.

3. Depending of the LINUX OS version installed it may be required to modify the LD_LIBRARY_PATH environment variable to use the dynamic library libssw.so by one of the 2 following possibilities :

   • Export the path or the directory containing the library LD_LIBRARY_PATH=path_of_the_library
   • For a definitive inclusion edit /etc/ld.so.conf and add the path of the lib directory. Then, update the cache by using /sbin/ldconfig as root

4. In a python script or in a interactive interpreter the main class can be imported with : from ssw_wrap import Aligner

5. Instantiate the Aligner class with initial parameters, including the reference subject sequence. Aligner(myref, match=2, mismatch=2, gap_open=3, gap_extension=1, report_secondary=False, report_cigar=False)

6. Call the object align method with a query sequence, the minimal score and length for the alignment to be reported res = ssw.align(myquery, min_score=10, min_len=20)

7. Parse the returned PyAlignRes object for alignment result description

###Run pyssw standalone

Usage: pyssw.py -s subject.fasta -q fastq (or fasta) [Facultative options]

Options:
 --version             show program's version number and exit
 -h, --help            show this help message and exit
 -s SUBJECT, --subject=SUBJECT Path of the fasta file containing the subject genome sequence [REQUIRED]
 -q QUERY, --query=QUERY   Path of the fastq or fasta file containing the short read to be aligned [REQUIRED]
 -t QTYPE, --qtype=QTYPE   Type of the query file = fastq or fasta. [default:fastq]
 -m MATCH, --match=MATCH   positive integer for weight match in genome sequence alignment. [default: 2]
 -x MISMATCH, --mismatch=MISMATCH  positive integer. The negative value will be used as weight mismatch in genome sequence alignment.[default: 2]
 -o GAP_OPEN, --gap_open=GAP_OPEN  positive integer. The negative value will be used as weight for the gap opening. [default: 3]
 -e GAP_EXTEND, --gap_extend=GAP_EXTEND    positive integer. The negative value will be used as weight for the gap opening. [default: 1]
 -f MIN_SCORE, --min_score=MIN_SCORE   integer. Consider alignments having a score <= MIN_SCORE as not aligned. [default: 0]
 -l MIN_LEN, --min_len=MIN_LEN integer. Consider alignments having a length <= as not aligned. [default: 0]
 -r, --reverse         Flag. Align query in forward and reverse orientation and choose the best alignment. [Set by default]
 -u, --unaligned       Flag. Write unaligned reads in sam output [Unset by default]

###pyssw output

For now, the program only output a SAM like file. The file encoding is sightly more complete than the version created by the C/C++ software and is fully compatible with downstream NGS utilities such as samtools. The MAPQ score field is set to 0, but the ssw score is reported in the AS optional tag.

@HD	VN:1.0	SO:unsorted
@SQ	SN:Virus_genome	LN:2216
@PG	ID:Striped-Smith-Waterman	PN:pyssw	VN:0.1
@CO	Score_values = match 2, mismatch 2, gap_open 3, gap_extend 1
@CO	Filter Options = min_score 250, min_len 0
M00785:2:000000000-A60JC:1:1101:14942:1516	16	Virus_genome	1773	0	151M	*	0	0    TTGTTT...TATTAC >A1AAF...@1@@2@	    AS:i:302
M00785:2:000000000-A60JC:1:1101:13644:1543	0	Virus_genome	1985	0	21M1I129M	*	0	0    TTTAAA...CTCGCT	AAAA11.../<</</     AS:i:289
M00785:2:000000000-A60JC:1:1101:13883:1547	16	Virus_genome	716	0	36M2D115M	*	0	0	CTTACC...GGAATT	>AA1>C...////?F    AS:i:294
...

###Speed and memory usage of pyssw

Intel Core i5 3320PM CPU @ 2.60GHz x 2

• E. coli K12 U0096.3 (4,641,652 nucleotides) vs 10,000 illumina reads (50pb) from CLoVR public repository

  pyssw.py -s Ecoli_K12_U0096.3.fa -q Ecoli_illumina_10k.fastq -r -f 80

  Time : ~55m 40s

• Short virus reference (2,216 nt) vs 100,000 illumina Miseq reads (150pb) (see demo dataset)

  pyssw.py -s Virus_genome.fa.gz -q 100k_illumina1.fastq.gz -f 250

  Time : ~60s

###Please cite this paper, if you need: http://dx.plos.org/10.1371/journal.pone.0082138