NEWS

Release 0.6.2 (19 June, 2012)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This is largely a bug-fix release. Notable changes in BWA-short and BWA-SW:

 * Bugfix: BWA-SW may give bad alignments due to incorrect band width.

 * Bugfix: A segmentation fault due to an out-of-boundary error. The fix is a
   temporary solution. The real cause has not been identified.

 * Attempt to read index from prefix.64.bwt, such that the 32-bit and 64-bit
   index can coexist.

 * Added options '-I' and '-S' to control BWA-SW pairing.

(0.6.2: 19 June 2012, r126)



Release 0.6.1 (28 November, 2011)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes to BWA-short:

 * Bugfix: duplicated alternative hits in the XA tag.

 * Bugfix: when trimming enabled, bwa-aln trims 1bp less.

 * Disabled the color-space alignment. 0.6.x is not working with SOLiD reads at
   present.

Notable changes to BWA-SW:

 * Bugfix: segfault due to excessive ambiguous bases.

 * Bugfix: incorrect mate position in the SE mode.

 * Bugfix: rare segfault in the PE mode

 * When macro _NO_SSE2 is in use, fall back to the standard Smith-Waterman
   instead of SSE2-SW.

 * Optionally mark split hits with lower alignment scores as secondary.

Changes to fastmap:

 * Bugfix: infinite loop caused by ambiguous bases.

 * Optionally output the query sequence.

(0.6.1: 28 November 2011, r104)



Release 0.5.10 and 0.6.0 (12 November, 2011)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

The 0.6.0 release comes with two major changes. Firstly, the index data
structure has been changed to support genomes longer than 4GB. The forward and
reverse backward genome is now integrated in one index. This change speeds up
BWA-short by about 20% and BWA-SW by 90% with the mapping acccuracy largely
unchanged. A tradeoff is BWA requires more memory, but this is the price almost
all mappers that index the genome have to pay.

Secondly, BWA-SW in 0.6.0 now works with paired-end data. It is more accurate
for highly unique reads and more robust to long indels and structural
variations. However, BWA-short still has edges for reads with many suboptimal
hits. It is yet to know which algorithm is the best for variant calling.

0.5.10 is a bugfix release only and is likely to be the last release in the 0.5
branch unless I find critical bugs in future.

Other notable changes:

 * Added the `fastmap' command that finds super-maximal exact matches. It does
   not give the final alignment, but runs much faster. It can be a building
   block for other alignment algorithms. [0.6.0 only]

 * Output the timing information before BWA exits. This also tells users that
   the task has been finished instead of being killed or aborted. [0.6.0 only]

 * Sped up multi-threading when using many (>20) CPU cores.

 * Check I/O error.

 * Increased the maximum barcode length to 63bp.

 * Automatically choose the indexing algorithm.

 * Bugfix: very rare segfault due to an uninitialized variable. The bug also
   affects the placement of suboptimal alignments. The effect is very minor.

This release involves quite a lot of tricky changes. Although it has been
tested on a few data sets, subtle bugs may be still hidden. It is *NOT*
recommended to use this release in a production pipeline. In future, however,
BWA-SW may be better when reads continue to go longer. I would encourage users
to try the 0.6 release. I would also like to hear the users' experience. Thank
you.

(0.6.0: 12 November 2011, r85)



Beta Release 0.5.9 (24 January, 2011)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes:

 * Feature: barcode support via the `-B' option.

 * Feature: Illumina 1.3+ read format support via the `-I' option.

 * Bugfix: RG tags are not attached to unmapped reads.

 * Bugfix: very rare bwasw mismappings

 * Recommend options for PacBio reads in bwasw help message.


Also, since January 13, the BWA master repository has been moved to github:

  https://github.com/lh3/bwa

The revision number has been reset. All recent changes will be first
committed to this repository.

(0.5.9: 24 January 2011, r16)



Beta Release Candidate 0.5.9rc1 (10 December, 2010)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes in bwasw:

 * Output unmapped reads.

 * For a repetitive read, choose a random hit instead of a fixed
   one. This is not well tested.

Notable changes in bwa-short:

 * Fixed a bug in the SW scoring system, which may lead to unexpected
   gaps towards the end of a read.

 * Fixed a bug which invalidates the randomness of repetitive reads.

 * Fixed a rare memory leak.

 * Allowed to specify the read group at the command line.

 * Take name-grouped BAM files as input.

Changes to this release are usually safe in that they do not interfere
with the key functionality. However, the release has only been tested on
small samples instead of on large-scale real data. If anything weird
happens, please report the bugs to the bio-bwa-help mailing list.

(0.5.9rc1: 10 December 2010, r1561)



Beta Release 0.5.8 (8 June, 2010)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes in bwasw:

 * Fixed an issue of missing alignments. This should happen rarely and
   only when the contig/read alignment is multi-part. Very rarely, bwasw
   may still miss a segment in a multi-part alignment. This is difficult
   to fix, although possible.

Notable changes in bwa-short:

 * Discard the SW alignment when the best single-end alignment is much
   better. Such a SW alignment may caused by structural variations and
   forcing it to be aligned leads to false alignment. This fix has not
   been tested thoroughly. It would be great to receive more users
   feedbacks on this issue.

 * Fixed a typo/bug in sampe which leads to unnecessarily large memory
   usage in some cases.

 * Further reduced the chance of reporting `weird pairing'.

(0.5.8: 8 June 2010, r1442)



Beta Release 0.5.7 (1 March, 2010)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This release only has an effect on paired-end data with fat insert-size
distribution. Users are still recommended to update as the new release
improves the robustness to poor data.

 * The fix for `weird pairing' was not working in version 0.5.6, pointed
   out by Carol Scott. It should work now.

 * Optionally output to a normal file rather than to stdout (by Tim
   Fennel).

(0.5.7: 1 March 2010, r1310)



Beta Release 0.5.6 (10 Feburary, 2010)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes in bwa-short:

 * Report multiple hits in the SAM format at a new tag XA encoded as:
   (chr,pos,CIGAR,NM;)*. By default, if a paired or single-end read has
   4 or fewer hits, they will all be reported; if a read in a anomalous
   pair has 11 or fewer hits, all of them will be reported.

 * Perform Smith-Waterman alignment also for anomalous read pairs when
   both ends have quality higher than 17. This reduces false positives
   for some SV discovery algorithms.

 * Do not report "weird pairing" when the insert size distribution is
   too fat or has a mean close to zero.

 * If a read is bridging two adjacent chromsomes, flag it as unmapped.

 * Fixed a small but long existing memory leak in paired-end mapping.

 * Multiple bug fixes in SOLiD mapping: a) quality "-1" can be correctly
   parsed by solid2fastq.pl; b) truncated quality string is resolved; c)
   SOLiD read mapped to the reverse strand is complemented.

 * Bwa now calculates skewness and kurtosis of the insert size
   distribution.

 * Deploy a Bayesian method to estimate the maximum distance for a read
   pair considered to be paired properly. The method is proposed by
   Gerton Lunter, but bwa only implements a simplified version.

 * Export more functions for Java bindings, by Matt Hanna (See:
   http://www.broadinstitute.org/gsa/wiki/index.php/Sting_BWA/C_bindings)

 * Abstract bwa CIGAR for further extension, by Rodrigo Goya.

(0.5.6: 10 Feburary 2010, r1303)



Beta Release 0.5.5 (10 November, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This is a bug fix release:

 * Fixed a serious bug/typo in aln which does not occur given short
   reads, but will lead to segfault for >500bp reads. Of course, the aln
   command is not recommended for reads longer than 200bp, but this is a
   bug anyway.

 * Fixed a minor bug/typo which leads to incorrect single-end mapping
   quality when one end is moved to meet the mate-pair requirement.

 * Fixed a bug in samse for mapping in the color space. This bug is
   caused by quality filtration added since 0.5.1.

(0.5.5: 10 November 2009, r1273)



Beta Release 0.5.4 (9 October, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Since this version, the default seed length used in the "aln" command is
changed to 32.

Notable changes in bwa-short:

 * Added a new tag "XC:i" which gives the length of clipped reads.

 * In sampe, skip alignments in case of a bug in the Smith-Waterman
   alignment module.

 * In sampe, fixed a bug in pairing when the read sequence is identical
   to its reverse complement.

 * In sampe, optionally preload the entire FM-index into memory to
   reduce disk operations.

Notable changes in dBWT-SW/BWA-SW:

 * Changed name dBWT-SW to BWA-SW.

 * Optionally use "hard clipping" in the SAM output.

(0.5.4: 9 October 2009, r1245)



Beta Release 0.5.3 (15 September, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Fixed a critical bug in bwa-short: reads mapped to the reverse strand
are not complemented.

(0.5.3: 15 September 2009, r1225)



Beta Release 0.5.2 (13 September, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes in bwa-short:

 * Optionally trim reads before alignment. See the manual page on `aln
   -q' for detailed description.

 * Fixed a bug in calculating the NM tag for a gapped alignment.

 * Fixed a bug given a mixture of reads with some longer than the seed
   length and some shorter.

 * Print SAM header.

Notable changes in dBWT-SW:

 * Changed the default value of -T to 30. As a result, the accuracy is a
   little higher for short reads at the cost of speed.

(0.5.2: 13 September 2009, r1223)



Beta Release 0.5.1 (2 September, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes in the short read alignment component:

 * Fixed a bug in samse: do not write mate coordinates.

Notable changes in dBWT-SW:

 * Randomly choose one alignment if the read is a repetitive.

 * Fixed a flaw when a read is mapped across two adjacent reference
   sequences. However, wrong alignment reports may still occur rarely in
   this case.

 * Changed the default band width to 50. The speed is slower due to this
   change.

 * Improved the mapping quality a little given long query sequences.

(0.5.1: 2 September 2009, r1209)



Beta Release 0.5.0 (20 August, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This release implements a novel algorithm, dBWT-SW, specifically
designed for long reads. It is 10-50 times faster than SSAHA2, depending
on the characteristics of the input data, and achieves comparable
alignment accuracy while allowing chimera detection. In comparison to
BLAT, dBWT-SW is several times faster and much more accurate especially
when the error rate is high. Please read the manual page for more
information.

The dBWT-SW algorithm is kind of developed for future sequencing
technologies which produce much longer reads with a little higher error
rate. It is still at its early development stage. Some features are
missing and it may be buggy although I have evaluated on several
simulated and real data sets. But following the "release early"
paradigm, I would like the users to try it first.

Other notable changes in BWA are:

 * Fixed a rare bug in the Smith-Waterman alignment module.

 * Fixed a rare bug about the wrong alignment coordinate when a read is
   poorly aligned.

 * Fixed a bug in generating the "mate-unmap" SAM tag when both ends in
   a pair are unmapped.

(0.5.0: 20 August 2009, r1200)



Beta Release 0.4.9 (19 May, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Interestingly, the integer overflow bug claimed to be fixed in 0.4.7 has
not in fact. Now I have fixed the bug. Sorry for this and thank Quan
Long for pointing out the bug (again).

(0.4.9: 19 May 2009, r1075)



Beta Release 0.4.8 (18 May, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

One change to "aln -R". Now by default, if there are no more than `-R'
equally best hits, bwa will search for suboptimal hits. This change
affects the ability in finding SNPs in segmental duplications.

I have not tested this option thoroughly, but this simple change is less
likely to cause new bugs. Hope I am right.

(0.4.8: 18 May 2009, r1073)



Beta Release 0.4.7 (12 May, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes:

 * Output SM (single-end mapping quality) and AM (smaller mapping
   quality among the two ends) tag from sam output.

 * Improved the functionality of stdsw.

 * Made the XN tag more accurate.

 * Fixed a very rare segfault caused by integer overflow.

 * Improve the insert size estimation.

 * Fixed compiling errors for some Linux systems.

(0.4.7: 12 May 2009, r1066)



Beta Release 0.4.6 (9 March, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This release improves the SOLiD support. First, a script for converting
SOLiD raw data is provided. This script is adapted from solid2fastq.pl
in the MAQ package. Second, a nucleotide reference file can be directly
used with `bwa index'. Third, SOLiD paired-end support is
completed. Fourth, color-space reads will be converted to nucleotides
when SAM output is generated. Color errors are corrected in this
process. Please note that like MAQ, BWA cannot make use of the primer
base and the first color.

In addition, the calculation of mapping quality is also improved a
little bit, although end-users may barely observe the difference.

(0.4.6: 9 March 2009, r915)



Beta Release 0.4.5 (18 Feburary, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Not much happened, but I think it would be good to let the users use the
latest version.

Notable changes (Thank Bob Handsaker for catching the two bugs):

 * Improved bounary check. Previous version may still give incorrect
   alignment coordinates in rare cases.

 * Fixed a bug in SW alignment when no residue matches. This only
   affects the `sampe' command.

 * Robustly estimate insert size without setting the maximum on the
   command line. Since this release `sampe -a' only has an effect if
   there are not enough good pairs to infer the insert size
   distribution.

 * Reduced false PE alignments a little bit by using the inferred insert
   size distribution. This fix may be more important for long insert
   size libraries.

(0.4.5: 18 Feburary 2009, r829)



Beta Release 0.4.4 (15 Feburary, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This is mainly a bug fix release. Notable changes are:

 * Imposed boundary check for extracting subsequence from the
   genome. Previously this causes memory problem in rare cases.

 * Fixed a bug in failing to find whether an alignment overlapping with
   N on the genome.

 * Changed MD tag to meet the latest SAM specification.

(0.4.4: 15 Feburary 2009, r815)



Beta Release 0.4.3 (22 January, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Notable changes:

 * Treat an ambiguous base N as a mismatch. Previous versions will not
   map reads containing any N.

 * Automatically choose the maximum allowed number of differences. This
   is important when reads of different lengths are mixed together.

 * Print mate coordinate if only one end is unmapped.

 * Generate MD tag. This tag encodes the mismatching positions and the
   reference bases at these positions. Deletions from the reference will
   also be printed.

 * Optionally dump multiple hits from samse, in another concise format
   rather than SAM.

 * Optionally disable iterative search. This is VERY SLOOOOW, though.

 * Fixed a bug in generate SAM.

(0.4.3: 22 January 2009, r787)



Beta Release 0.4.2 (9 January, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Aaron Quinlan found a bug in the indexer: the bwa indexer segfaults if
there are no comment texts in the FASTA header. This is a critical
bug. Nothing else was changed.

(0.4.2: 9 January 2009, r769)



Beta Release 0.4.1 (7 January, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

I am sorry for the quick updates these days. I like to set a milestone
for BWA and this release seems to be. For paired end reads, BWA also
does Smith-Waterman alignment for an unmapped read whose mate can be
mapped confidently. With this strategy BWA achieves similar accuracy to
maq. Benchmark is also updated accordingly.

(0.4.1: 7 January 2009, r760)



Beta Release 0.4.0 (6 January, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

In comparison to the release two days ago, this release is mainly tuned
for performance with some tricks I learnt from Bowtie. However, as the
indexing format has also been changed, I have to increase the version
number to 0.4.0 to emphasize that *DATABASE MUST BE RE-INDEXED* with
`bwa index'.

 * Improved the speed by about 20%.

 * Added multi-threading to `bwa aln'.

(0.4.0: 6 January 2009, r756)



Beta Release 0.3.0 (4 January, 2009)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 * Added paired-end support by separating SA calculation and alignment
   output.

 * Added SAM output.

 * Added evaluation to the documentation.

(0.3.0: 4 January 2009, r741)



Beta Release 0.2.0 (15 Augusst, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 * Take the subsequence at the 5'-end as seed. Seeding strategy greatly
   improves the speed for long reads, at the cost of missing a few true
   hits that contain many differences in the seed. Seeding also increase
   the memory by 800MB.

 * Fixed a bug which may miss some gapped alignments. Fixing the bug
   also slows the speed a little.

(0.2.0: 15 August 2008, r428)



Beta Release 0.1.6 (08 Augusst, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 * Give accurate CIGAR string.

 * Add a simple interface to SW/NW alignment

(0.1.6: 08 August 2008, r414)



Beta Release 0.1.5 (27 July, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 * Improve the speed. This version is expected to give the same results.

(0.1.5: 27 July 2008, r400)



Beta Release 0.1.4 (22 July, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

 * Fixed a bug which may cause missing gapped alignments.

 * More clearly define what alignments can be found by BWA (See
   manual). Now BWA runs a little slower because it will visit more
   potential gapped alignments.

 * A bit code clean up.

(0.1.4: 22 July 2008, r387)



Beta Release 0.1.3 (21 July, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Improve the speed with some tricks on retrieving occurences. The results
should be exactly the same as that of 0.1.2.

(0.1.3: 21 July 2008, r382)



Beta Release 0.1.2 (17 July, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Support gapped alignment. Codes for ungapped alignment has been removed.

(0.1.2: 17 July 2008, r371)



Beta Release 0.1.1 (03 June, 2008)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This is the first release of BWA, Burrows-Wheeler Alignment tool. Please
read man page for more information about this software.

(0.1.1: 03 June 2008, r349)