- Please cite: Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 10:R25.
- For release updates, subscribe to the mailing list.
All indexes are for assemblies, not contigs. Unplaced or unlocalized sequences and alternate haplotype assemblies are excluded.
Some unzip programs cannot handle archives >2 GB. If you have problems downloading or unzipping a >2 GB index, try downloading in two parts.
Check .zip file integrity with MD5s.
- Langmead B, Trapnell C, Pop M, Salzberg SL. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biology 10:R25.
- Langmead B, Schatz M, Lin J, Pop M, Salzberg SL. Searching for SNPs with cloud computing. Genome Biology 10:R134.
- Trapnell C, Pachter L, Salzberg SL, TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 2009 25(9):1105-1111.
1.1.0 - 7/19/2014
- Added support for large and small indexes, removing 4-billion-nucleotide barrier. Bowtie can now be used with reference genomes of any size.
- No longer releasing 32-bit binaries. Simplified manual and Makefile accordingly.
- Phased out CygWin support.
- Improved efficiency of index files loading.
- Fixed a bug that made bowtie-inspect fail in some situations.
- (This release was briefly given version number 2.0.0, but we changed it to 1.1.0 to avoid confusion with Bowtie 2.)
1.0.1 release - 3/14/2014
- Improved index querying efficiency using "population count" instructions available since SSE4.2.
- Credits to the Intel(r) enabling team for performance optimizations included in this release. Thank you!
Bowtie on GitHub - 4/11/13
- Bowtie source now lives in a public GitHub repository.
1.0.0 release - 4/9/13
- Finally, a 64-bit Windows binary!
- Due to general performance improvements spinlocking is now used by default. The EXTRA_FLAGS=-DNO_SPINLOCK may be used to reverse this during compilation.
- Fixed some race conditions.
- CygWin builds now use pthreads library.
- Changed MinGW to optionally use pthread library on Win32 platforms.
- Changed the Windows build to use native Windows threads by default.
- Renamed COPYING to LICENSE to be more GitHub-friendly.
- Tokenizer no longer has limit of 10,000 tokens, which was a problem for users trying to index a very large number of FASTA files.
- Removed references to no-longer-implemented --cutoff option for bowtie-build
0.12.9 release - 12/16/12
- Fixed a bug whereby read names would not be truncated at first whitespace character in unmapped or maxed-out SAM records.
- Fixed errors and warnings when compiling with clang++.
- Fixed most errors and warnings when compiling with recent versions of g++, though you may need to add EXTRA_FLAGS=-Wno-enum-compare to avoid all warnings.
0.12.8 release - 5/6/12
- Fixed a bug that would sometimes cause an immediate segmentation fault in --sam mode.
- Fixed make_galGal3.sh script to not omit chromosome 25.
- Removed -B option from usage message for bowtie-build; that option is not implemented.
- Fixed issue that could cause Bowtie not to compile when BOWTIE_PTHREADS is left undefined and pthreads.h is not present.
- Elaborated documentation for -B/--offbase option to indicate that it only affects offsets in the default output mode, not in SAM mode.
Bowtie 2 beta released - 10/16/2011
- Bowtie 2 2.0.0-beta2 is available now.
- Differences between Bowtie 2 and Bowtie 1 include:
- For reads longer than about 50 bp Bowtie 2 is generally faster, more sensitive, and uses less memory than Bowtie 1. For relatively short reads (e.g. less than 50 bp) Bowtie 1 is sometimes faster and/or more sensitive.
- Bowtie 2 supports gapped alignment with affine gap penalties. Number of gaps and gap lengths are not restricted, except by way of the configurable scoring scheme. Bowtie 1 finds just ungapped alignments.
- Bowtie 2 supports local alignment, which doesn't require reads to align end-to-end. Local alignments might be "trimmed" ("soft clipped") at one or both extremes in a way that optimizes alignment score. Bowtie 2 also supports end-to-end alignment which, like Bowtie 1, requires that the read align entirely.
- There is no upper limit on read length in Bowtie 2. Bowtie 1 has an upper limit of around 1000 bp.
- Bowtie 2 allows alignments to overlap ambiguous characters (e.g. Ns) in the reference. Bowtie 1 does not.
- Bowtie 2 does away with Bowtie 1's notion of alignment "stratum", and its distinction between "Maq-like" and "end-to-end" modes. In Bowtie 2 all alignments lie along a continuous spectrum of alignment scores where the scoring scheme, similar to Needleman-Wunsch and Smith-Waterman.
- Bowtie 2's paired-end alignment is more flexible. E.g. for pairs that do not align in a paired fashion, Bowtie 2 attempts to find unpaired alignments for each mate.
- Bowtie 2 does not align colorspace reads.
Fixed chicken index - 7/19/2011
- The pre-built chicken genome available from this website was missing chr25. chr25_random was included, but chr25 was erroneously excluded. This is fixed as of today - the index files linked to from the sidebar now contains all chicken chromosomes.
0.12.7 release - 9/7/10
- Fixes the all-gap reference sequence issue that was present in Bowtie 0.12.6. Index files produced by bowtie-build 0.12.5 and earlier (back to 0.10.*) are compatible with bowtie 0.12.7. Index files produced by bowtie-build 0.12.7 are backward compatible with bowtie 0.12.5 and earlier as long as the first reference sequence is not all-gaps (or, when colorspace indexes, as long as the first reference sequence has two consecutive ACGT characters in it somewhere).
- Indexes where the first sequence consists of all gaps or other non-ACGT characters are not handled properly by Bowtie versions 0.12.5 and older, but are handled properly by Bowtie 0.12.7.
- REMOVED: bowtie-build's --old-reverse option; the old reverse- index scheme is again the default. The new scheme is disabled pending further refinement.
0.12.6 release - 8/29/10
- Modified bowtie-inspect's default mode to use the bit-encoded reference portion of the index to reconstruct the reference sequence, rather than the ebwt portion. This makes bowtie-inspect much faster and uses less memory, and the output for a colorspace index will now be in nucleotide space. To get the behavior of the old default, use the new -e/--ebwt-ref option.
- Fixed bug whereby SOLiD QV strings without a trailing space would fail to parse.
- Moved to a new default way of building the reverse index. Revert to the old behavior with bowtie-build's new --old-reverse option. The new reverse index format is forward and backward compatible with `bowtie`, unless otherwise noted in a future version.
- Fixed issue that would sometimes cause bowtie-build to crash when building a large index with a low --offrate.
- Fixed build issue that would cause bowtie-build built on one version of Linux to die with a "floating point error" on other versions. bowtie now simply skips reads with 0 characters. Previously it would print an error and exit.
- Fixed a bug whereby alignment cost could sometimes be miscalculated. Stratum was unaffected.
- bowtie now simply skips reads with 0 characters. Previously it would print an error and exit.
Myrna paper out - 8/11/10
- The provisional version of the Myrna paper appeared today in Genome Biology. See the Myrna site for details about the software.
Myrna released - 7/22/10
- Myrna is cloud-scale tool for calculating differential gene expression in large RNA-seq datasets. It is very similar to Crossbow in design and philosophy; e.g. it uses Bowtie, it's "tri-mode" (runs in the cloud or on a Hadoop cluster or a single computer), it has a web-form-based GUI for composing and submitting cloud jobs, and its cloud mode uses Elastic MapReduce.
Major update to Crossbow - 7/22/10
- Crossbow has received a major update to
version 1.0.4. New features:
- New "tri-mode" design; Crossbow jobs can be run in the cloud (on Amazon AWS), on a Hadoop cluster, or on a single computer, exploiting multiple processor cores
- Added a web-form-based GUI for composing and submitting cloud jobs.
- Cloud version of Crossbow now runs on Elastic MapReduce (EMR) rather than directly on Elastic Compute Cloud (EC2).