Beta Release 0.1.16 (21 April, 2011)
Notable changes in samtools:
* Support the new SAM/BAM type `B' in the latest SAM spec v1.4.
* When the output file of `samtools merge' exists, do not overwrite it unless
a new command-line option `-f' is applied.
* Bugfix: BED support is not working when the input BED is not sorted.
* Bugfix: some reads without coordinates but given on the reverse strand are
lost in merging.
Notable changes in bcftools:
* Code cleanup: separated max-likelihood inference and Bayesian inference.
* Test Hardy-Weinberg equilibrium with a likelihood-ratio test.
* Provided another association test P-value by likelihood-ratio test.
* Use Brent's method to estimate the site allele frequency when EM converges
slowly. The resulting ML estimate of allele frequnecy is more accurate.
* Added the `ldpair' command, which computes r^2 between SNP pairs given in
an input file.
Also, the `pileup' command, which has been deprecated by `mpileup' since
version 0.1.10, will be dropped in the next release. The old `pileup' command
is substandard and causing a lot of confusion.
(0.1.16: 21 April 2011, r963:234)
Beta Release 0.1.15 (10 April, 2011)
* Allow to perform variant calling or to extract information in multiple
regions specified by a BED file (`samtools mpileup -l', `samtools view -L'
and `bcftools view -l').
* Added the `depth' command to samtools to compute the per-base depth with a
simpler interface. File `bam2depth.c', which implements this command, is the
recommended example on how to use the mpileup APIs.
* Estimate genotype frequencies with ML; perform chi^2 based Hardy-Weinberg
test using this estimate.
* For `samtools view', when `-R' is specified, drop read groups in the header
that are not contained in the specified file.
* For `samtools flagstat', separate QC-pass and QC-fail reads.
* Improved the command line help of `samtools mpileup' and `bcftools view'.
* Use a global variable to control the verbose level of samtools stderr
output. Nonetheless, it has not been full utilized.
* Fixed an issue in association test which may report false associations,
possibly due to floating point underflow.
(0.1.15: 10 April 2011, r949:203)
Beta release 0.1.14 (21 March, 2011)
This release implements a method for testing associations for case-control
data. The method does not call genotypes but instead sums over all genotype
configurations to compute a chi^2 based test statistics. It can be potentially
applied to comparing a pair of samples (e.g. a tumor-normal pair), but this
has not been evaluated on real data.
Another new feature is to make X chromosome variant calls when female and male
samples are both present. The user needs to provide a file indicating the
ploidy of each sample (see also manual bcftools/bcftools.1).
Other notable changes:
* Added `bcftools view -F' to parse BCF files generated by samtools r921 or
older which encodes PL in a different way.
* Changed the behavior of `bcftools view -s'. Now when a list of samples is
provided, the samples in the output will be reordered to match the ordering
in the sample list. This change is mainly designed for association test.
* Sped up `bcftools view -v' for target sequencing given thousands of samples.
Also added a new option `view -d' to skip loci where only a few samples are
covered by reads.
* Dropped HWE test. This feature has never been implemented properly. An EM
should be much better. To be implemented in future.
* Added the `cat' command to samtools. This command concatenate BAMs with
identical sequence dictionaries in an efficient way. Modified from bam_cat.c
written by Chris Saunders.
* Added `samtools view -1' to write BAMs at a low compression level but twice
faster to create. The `sort' command generates temporary files at a low
compression level as well.
* Added `samtools mpileup -6' to accept "BAM" with Illumina 1.3+ quality
strings (strictly speaking, such a file is not BAM).
* Added `samtools mpileup -L' to skip INDEL calling in regions with
excessively high coverage. Such regions dramatically slow down mpileup.
* Updated `misc/export2sam.pl', provided by Chris Saunders from Illumina Inc.
(0.1.14: 21 March 2011, r933:170)
Beta release 0.1.13 (1 March, 2011)
The most important though largely invisible modification is the change of the
order of genotypes in the PL VCF/BCF tag. This is to conform the upcoming VCF
spec v4.1. The change means that 0.1.13 is not backward compatible with VCF/BCF
generated by samtools older than r921 inclusive. VCF/BCF generated by the new
samtools will contain a line `##fileformat=VCFv4.1' as well as the samtools
Single Individual Haplotyping (SIH) is added as an experimental feature. It
originally aims to produce haploid consensus from fosmid pool sequencing, but
also works with short-read data. For short reads, phased blocks are usually too
short to be useful in many applications, but they can help to rule out part of
SNPs close to INDELs or between copies of CNVs.
Other notable changes in samtools:
* Construct per-sample consensus to reduce the effect of nearby SNPs in INDEL
calling. This reduces the power but improves specificity.
* Improved sorting order checking in indexing. Now indexing is the preferred way
to check if a BAM is sorted.
* Added a switch `-E' to mpileup and calmd. This option uses an alternative way
to apply BAQ, which increases sensistivity, especially to MNPs, at the cost of
a little loss in specificity.
* Added `mpileup -A' to allow to use reads in anomalous pairs in SNP calling.
* Added `mpileup -m' to allow fine control of the collection of INDEL candidates.
* Added `mpileup -S' to compute per-sample strand bias P-value.
* Added `mpileup -G' to exclude read groups in variant calling.
* Fixed segfault in indel calling related to unmapped and refskip reads.
* Fixed an integer overflow in INDEL calling. This bug produces wrong INDEL
genotypes for longer short INDELs, typically over 10bp.
* Fixed a bug in tview on big-endian machines.
* Fixed a very rare memory issue in bam_md.c
* Fixed an out-of-boundary bug in mpileup when the read base is `N'.
* Fixed a compiling error when the knetfile library is not used. Fixed a
library compiling error due to the lack of bam_nt16_nt4_table table.
Suppress a compiling warning related to the latest zlib.
Other notable changes in bcftools:
* Updated the BCF spec.
* Added the `FQ' VCF INFO field, which gives the phred-scaled probability
of all samples being the same (identical to the reference or all homozygous
variants). Option `view -f' has been dropped.
* Implementated of "vcfutils.pl vcf2fq" to generate a consensus sequence
similar to "samtools.pl pileup2fq".
* Make sure the GT FORMAT field is always the first FORMAT to conform the VCF
spec. Drop bcf-fix.pl.
* Output bcftools specific INFO and FORMAT in the VCF header.
* Added `view -s' to call variants from a subset of samples.
* Properly convert VCF to BCF with a user provided sequence dictionary. Nonetheless,
custom fields are still unparsed and will be stored as a missing value.
* Fixed a minor bug in Fisher's exact test; the results are rarely changed.
(0.1.13: 1 March 2011, r926:134)
Beta release 0.1.12a (2 December, 2010)
This is another bug fix release:
* Fixed a memory violation in mpileup, which causes segfault. Release
0.1.9 and above are affected.
* Fixed a memory violation in the indel caller, which does not causes
segfault, but may potentially affect deletion calls in an unexpected
way. Release 0.1.10 and above are affected.
* Fixed a bug in computing r-square in bcftools. Few are using this
functionality and it only has minor effect.
* Fixed a memory leak in bam_fetch().
* Fixed a bug in writing meta information to the BAM index for the last
sequence. This bug is invisible to most users, but it is a bug anyway.
* Fixed a bug in bcftools which causes false "DP4=0,0,0,0" annotations.
(0.1.12: 2 December 2010, r862)
Beta release 0.1.11 (21 November, 2010)
This is mainly a bug fix release:
* Fixed a bug in random retrieval (since 0.1.8). It occurs when reads
are retrieved from a small region containing no reads.
* Fixed a bug in pileup (since 0.1.9). The bug causes an assertion
failure when the first CIGAR operation is a deletion.
* Improved fault tolerence in remote access.
One minor feature has been implemented in bcftools:
* Added a reference-free variant calling mode. In this mode, a site is
regarded as a variat iff the sample(s) contains two or more alleles;
the meaning of the QUAL field in the VCF output is changed
accordingly. Effectively, the reference allele is irrelevant to the
result in the new mode, although the reference sequence has to be
used in realignment when SAMtools computes genotype likelihoods.
In addition, since 0.1.10, the `pileup' command has been deprecated by
`mpileup' which is more powerful and more accurate. The `pileup' command
will not be removed in the next few releases, but new features will not
(0.1.11: 21 November 2010, r851)
Beta Release 0.1.10 (16 November, 2010)
This release is featured as the first major improvement to the indel
caller. The method is similar to the old one implemented in the pileup
command, but the details are handled more carefully both in theory and
in practice. As a result, the new indel caller usually gives more
accurate indel calls, though at the cost of sensitivity. The caller is
implemented in the mpileup command and is invoked by default. It works
with multiple samples.
Other notable changes:
* With the -r option, the calmd command writes the difference between
the original base quality and the BAQ capped base quality at the BQ
tag but does not modify the base quality. Please use -Ar to overwrite
the original base quality (the 0.1.9 behavior).
* Allow to set a maximum per-sample read depth to reduce memory. In
0.1.9, most of memory is wasted for the ultra high read depth in some
regions (e.g. the chr1 centromere).
* Optionally write per-sample read depth and per-sample strand bias
* Compute equal-tail (Bayesian) credible interval of site allele
frequency at the CI95 VCF annotation.
* Merged the vcfutils.pl varFilter and filter4vcf for better SNP/indel
(0.1.10: 16 November 2010, r829)
Beta Release 0.1.9 (27 October, 2010)
This release is featured as the first major improvement to the samtools'
SNP caller. It comes with a revised MAQ error model, the support of
multi-sample SNP calling and the computation of base alignment quality
The revised MAQ error model is based on the original model. It solves an
issue of miscalling SNPs in repetitive regions. Althought such SNPs can
usually be filtered at a later step, they mess up unfiltered calls. This
is a theoretical flaw in the original model. The revised MAQ model
deprecates the orginal MAQ model and the simplified SOAPsnp model.
Multi-sample SNP calling is separated in two steps. The first is done by
samtools mpileup and the second by a new program, bcftools, which is
included in the samtools source code tree. Multi-sample SNP calling also
works for single sample and has the advantage of enabling more powerful
filtration. It is likely to deprecate pileup in future once a proper
indel calling method is implemented.
BAQ is the Phred-scaled probability of a read base being wrongly
aligned. Capping base quality by BAQ has been shown to be very effective
in suppressing false SNPs caused by misalignments around indels or in
low-complexity regions with acceptable compromise on computation
time. This strategy is highly recommended and can be used with other SNP
callers as well.
In addition to the three major improvements, other notable changes are:
* Changes to the pileup format. A reference skip (the N CIGAR operator)
is shown as '<' or '>' depending on the strand. Tview is also changed
* Accelerated pileup. The plain pileup is about 50% faster.
* Regional merge. The merge command now accepts a new option to merge
files in a specified region.
* Fixed a bug in bgzip and razip which causes source files to be
deleted even if option -c is applied.
* In APIs, propogate errors to downstream callers and make samtools
return non-zero values once errors occur.
(0.1.9: 27 October 2010, r783)
Beta Release 0.1.8 (11 July, 2010)
Notable functional changes:
* Added the `reheader' command which replaces a BAM header with a new
header. This command is much faster than replacing header by
* Added the `mpileup' command which computes the pileup of multiple
* The `index' command now stores the number of mapped and unmapped
reads in the index file. This information can be retrieved quickly by
the new `idxstats' command.
* By default, pileup used the SOAPsnp model for SNP calling. This
avoids the floating overflow in the MAQ model which leads to spurious
calls in repetitive regions, although these calls will be immediately
filtered by varFilter.
* The `tview' command now correctly handles CIGARs like 7I10M and
10M1P1I10M which cause assertion failure in earlier versions.
* Tview accepts a region like `=10,000' where `=' stands for the
current sequence name. This saves typing for long sequence names.
* Added the `-d' option to `pileup' which avoids slow indel calling
in ultradeep regions by subsampling reads locally.
* Added the `-R' option to `view' which retrieves alignments in read
groups listed in the specified file.
* The BAM->SAM conversion is up to twice faster, depending on the
characteristic of the input.
* Parsing SAM headers with a lot of reference sequences is now much
* The number of lseek() calls per query is reduced when the query
region contains no read alignments.
* Fixed an issue in the indel caller that leads to miscall of indels.
Note that this solution may not work well when the sequencing indel
error rate is higher than the rate of SNPs.
* Fixed another issue in the indel caller which may lead to incorrect
* Fixed a bug in `sort' when option `-o' is applied.
* Fixed a bug in `view -r'.
APIs and other changes:
* Added iterator interfaces to random access and pileup. The callback
interfaces directly call the iterator interfaces.
* The BGZF blocks holding the BAM header are indepedent of alignment
BGZF blocks. Alignment records shorter than 64kB is guaranteed to be
fully contained in one BGZF block. This change is fully compatible
with the old version of samtools/picard.
Changes in other utilities:
* Updated export2sam.pl by Chris Saunders.
* Improved the sam2vcf.pl script.
* Added a Python version of varfilter.py by Aylwyn Scally.
(0.1.8: 11 July 2010, r613)
Beta Release 0.1.7 (10 November, 2009)
* Improved the indel caller in complex scenariors, in particular for
long reads. The indel caller is now able to make reasonable indel
calls from Craig Venter capillary reads.
* Rewrote single-end duplicate removal with improved
performance. Paired-end reads are not touched.
* Duplicate removal is now library aware. Samtools remove potential
PCR/optical dupliates inside a library rather than across libraries.
* SAM header is now fully parsed, although this functionality is not
used in merging and so on.
* In samtools merge, optionally take the input file name as RG-ID and
attach the RG tag to each alignment.
* Added FTP support in the RAZF library. RAZF-compressed reference
sequence can be retrieved remotely.
* Improved network support for Win32.
* Samtools sort and merge are now stable.
Changes in other utilities:
* Implemented sam2vcf.pl that converts the pileup format to the VCF
* This release of samtools is known to work with the latest
Bio-Samtools Perl module.
(0.1.7: 10 November 2009, r510)
Beta Release 0.1.6 (2 September, 2009)
* In tview, do not show a blank screen when no reads mapped to the
* Implemented native HTTP support in the BGZF library. Samtools is now
able to directly open a BAM file on HTTP. HTTP proxy is also
supported via the "http_proxy" environmental variable.
* Samtools is now compitable with the MinGW (win32) compiler and the
* The calmd (or fillmd) command now calculates the NM tag and replaces
MD tags if they are wrong.
* The view command now recognizes and optionally prints FLAG in HEXs or
strings to make a SAM file more friendly to human eyes. This is a
samtools-C extension, not implemented in Picard for the time
being. Please type `samtools view -?' for more information.
* BAM files now have an end-of-file (EOF) marker to facilitate
truncation detection. A warning will be given if an on-disk BAM file
does not have this marker. The warning will be seen on BAM files
generated by an older version of samtools. It does NO harm.
* New key bindings in tview: `r' to show read names and `s' to show
reference skip (N operation) as deletions.
* Fixed a bug in `samtools merge -n'.
* Samtools merge now optionally copies the header of a user specified
SAM file to the resultant BAM output.
* Samtools pileup/tview works with a CIGAR with the first or the last
operation is an indel.
* Fixed a bug in bam_aux_get().
Changes in other utilies:
* Fixed wrong FLAG in maq2sam.
(0.1.6: 2 September 2009, r453)
Beta Release 0.1.5 (7 July, 2009)
* Support opening a BAM alignment on FTP. Users can now use "tview" to
view alignments at the NCBI ftp site. Please read manual for more
* In library, propagate errors rather than exit or complain assertion
* Simplified the building system and fixed compiling errors caused by
* Fixed an issue about lost header information when a SAM is imported
with "view -t".
* Implemented "samtool.pl varFilter" which filters both SNPs and short
indels. This command replaces "indelFilter".
* Implemented "samtools.pl pileup2fq" to generate FASTQ consensus from
* In pileup, cap mapping quality at 60. This helps filtering when
different aligners are in use.
* In pileup, allow to output variant sites only.
* Made pileup generate correct calls in repetitive region. At the same
time, I am considering to implement a simplified model in SOAPsnp,
although this has not happened yet.
* In view, added '-u' option to output BAM without compression. This
option is preferred when the output is piped to other commands.
* In view, added '-l' and '-r' to get the alignments for one library or
read group. The "@RG" header lines are now partially parsed.
* Do not include command line utilities to libbam.a.
* Fixed memory leaks in pileup and bam_view1().
* Made faidx more tolerant to empty lines right before or after FASTA >
Changes in other utilities:
* Updated novo2sam.pl by Colin Hercus, the key developer of novoalign.
This release involves several modifications to the key code base which
may potentially introduce new bugs even though we have tried to minimize
this by testing on several examples. Please let us know if you catch
(0.1.5: 7 July 2009, r373)
Beta Release 0.1.4 (21 May, 2009)
* Added the 'rmdupse' command: removing duplicates for SE reads.
* Fixed a critical bug in the indel caller: clipped alignments are not
* Fixed a bug in the tview: gapped alignment may be incorrectly
* Unified the interface to BAM and SAM I/O. This is done by
implementing a wrapper on top of the old APIs and therefore old APIs
are still valid. The new I/O APIs also recognize the @SQ header
* Generate the MD tag.
* Generate "=" bases. However, the indel caller will not work when "="
bases are present.
* Enhanced support of color-read display (by Nils Homer).
* Implemented the GNU building system. However, currently the building
system does not generate libbam.a. We will improve this later. For
the time being, `make -f Makefile.generic' is preferred.
* Fixed a minor bug in pileup: the first read in a chromosome may be
* Fixed bugs in bam_aux.c. These bugs do not affect other components as
they were not used previously.
* Output the 'SM' tag from maq2sam.
(0.1.4: 21 May 2009, r297)
Beta Release 0.1.3 (15 April, 2009)
Notable changes in SAMtools:
* SAMtools is more consistent with the specification: a) '*' in the
QUAL field is allowed; b) the field separator is TAB only and SPACE
is treated as a character in a field; c) empty header is allowed.
* Implemented GLFv3 support in pileup.
* Fixed a severe bug in fixmate: strand information is wrongly
* Fixed a bug in alignment retrieval: alignments bridging n*16384bp are
not correctly retrieved sometimes.
* Fixed a bug in rmdup: segfault if unmapped reads are present.
* Move indel_filter.pl to samtools.pl and improved the filtering by
checking the actual number of alignments containing indels. The indel
pileup line is also changed a little to make this filtration easier.
* Fixed a minor bug in indexing: the bin number of an unmapped read is
* Added `flagstat' command to show statistics on the FLAG field.
* Improved indel caller by setting the maximum window size in local
Changes in other utilities:
* Fixed a bug in maq2sam: a tag name is obsolete.
* Improvement to wgsim: a) added support for SOLiD read simulation; b)
show the number of substitutions/indels/errors in read name; c)
considerable code clean up.
* Various converters: improved functionality in general.
* Updated the example SAM due to the previous bug in fixmate.
(0.1.3: 15 April 2009, r227)
Beta Release 0.1.2 (28 January, 2008)
Notable changes in SAMtools:
* Implemented a Bayesian indel caller. The new caller generate scores
and genotype and is potentially more accurate than Maq's indel
caller. The pileup format is also changed accordingly.
* Implemented rmdup command: remove potential PCR duplicates. Note that
this command ONLY works for FR orientation and requires ISIZE is
* Added fixmate command: fill in mate coordinates, ISIZE and mate
related flags from a name-sorted alignment.
* Fixed a bug in indexing: reads bridging 16x kbp were not retrieved.
* Allow to select reads shown in the pileup output with a mask.
* Generate GLFv2 from pileup.
* Added two more flags for flagging PCR/optical duplicates and for QC
* Fixed a bug in sort command: name sorting for large alignment did not
* Allow to completely disable RAZF (using Makefile.lite) as some people
have problem to compile it.
* Fixed a bug in import command when there are reads without
* Fixed a bug in tview: clipping broke the alignment viewer.
* Fixed a compiling error when _NO_CURSES is applied.
* Fixed a bug in merge command.
Changes in other utilities:
* Added wgsim, a paired-end reads simulator. Wgsim was adapted from
maq's reads simulator. Colin Hercus further improved it to allow
* Added wgsim_eval.pl, a script that evaluates the accuracy of
alignment on reads generated by wgsim.
* Added soap2sam.pl, a SOAP2->SAM converter. This converter does not
work properly when multiple hits are output.
* Added bowtie2sam.pl, a Bowtie->SAM converter. Only the top hit will
be retained when multiple hits are present.
* Fixed a bug in export2sam.pl for QC reads.
* Support RG tag at MAQ->SAM converter.
* Added novo2sam.pl, a NovoAlign->SAM converter. Multiple hits and
indel are not properly handled, though.
* Added zoom2sam.pl, a ZOOM->SAM converter. It only works with the
default Illumina output.
(0.1.2: 28 January 2008; r116)
Beta Release 0.1.1 (22 December, 2008)
The is the first public release of samtools. For more information,
please check the manual page `samtools.1' and the samtools website