SAM (Sequence Alignment/Map) format is a generic format for storing large nucleotide sequence alignments. SAM aims to be a format that:

SAM Tools provide various utilities for manipulating alignments in the SAM format, including sorting, merging, indexing and generating alignments in a per-position format.

SAMtools is hosted by The project page is here. The source code releases are available from the download page. You can check out the most recent source code from the github project page with:

  1. git clone git://







About the SAM/BAM Format

How can I get alignments in the SAM/BAM format?

Many alignment programs generate SAM/BAM natively or output a format that can be converted to SAM/BAM. Please check out this page for the complete list. If your preferred software is not in this list, you may contact the developers or write your own, and then please let us know.

How unaligned reads are stored in SAM?

An unaligned reads must be flagged with 0x4. It may have no coordinate (i.e. a coordinate `*:0'), but may have an ordinary coordinate with the CIGAR field set to `*'. In SAM, if one read in a read pair is aligned but the mate is not, we strongly recommend to set the coordinate of the unmapped read the same as that of the mapped one such that in a position sorted SAM/BAM file, the unmapped read is adjacent to the mapped. This convention greatly helps local assembly when we want to collect all related reads in a small region.

CIGAR is `50M', but I see mismatches in the alignment.

CIGAR operation `M' means `alignment match' (i.e. not a gap). It may be a `sequence match' or a `sequence mismatch'. Mismatching information is stored in the `MD' tag which is optional but can be generated with the `calmd' samtools command. We are proposing new CIGAR operations `=' for sequence match and `X' for sequence mismatch, but they are not well supported by samtools.

About SAMtools

How to convert SAM to BAM?

If your SAM file has header @SQ lines, you may get BAM by:

samtools view -bS aln.sam > aln.bam  

If not, you need to have your reference file ref.fa and then do this:

samtools faidx ref.fa  samtools view -bt ref.fa.fai aln.sam > aln.bam  

The second method also works if your SAM file has @SQ lines. After conversion, you would probably like to sort and index the alignment to enable fast random access:

samtools sort aln.bam aln-sorted  samtools index aln-sorted.bam  

I want to call SNPs and short indels.

For a short answer, do this:

samtools pileup -vcf ref.fa aln.bam | tee raw.txt | varFilter -D100 > flt.txt  awk '($3=="*"&&$6>=50)||($3!="*"&&$6>=20)' flt.txt > final.txt  

For a long answer, see this protocol. Please always remember to set the maximum depth (-D) in filtering.

I want to call SNPs from one chromosome only.

Index your alignment with the `index' command and:

samtools view -u aln.bam chr10 | samtools pileup -vcf ref.fa - > chr10.raw.txt  

Please read this page for more information.

The integer FLAG field is not friendly to eyes.

You may get string FLAG by:

samtools view -X aln.bam | less -S  

For more information, please check out:

samtools view -?  

I do not understand the columns in the pileup output.

This is explained in the manual page. Or briefly (when you invoke pileup with the -c option):

  1. reference sequence name
  2. reference coordinate
  3. reference base, or `*' for an indel line
  4. genotype where heterozygotes are encoded in the IUB code: M=A/C, R=A/G, W=A/T, S=C/G, Y=C/T and K=G/T; indels are indicated by, for example, */+A, -A/* or +CC/-C. There is no difference between */+A or +A/*.
  5. Phred-scaled likelihood that the genotype is wrong, which is also called `consensus quality'.
  6. Phred-scaled likelihood that the genotype is identical to the reference, which is also called `SNP quality'. Suppose the reference base is A and in alignment we see 17 G and 3 A. We will get a low consensus quality because it is difficult to distinguish an A/G heterozygote from a G/G homozygote. We will get a high SNP quality, though, because the evidence of a SNP is very strong.
  7. root mean square (RMS) mapping quality
  8. # reads covering the position
  9. read bases at a SNP line (check the manual page for more information); the 1st indel allele otherwise
  10. base quality at a SNP line; the 2nd indel allele otherwise
  11. indel line only: # reads directly supporting the 1st indel allele
  12. indel line only: # reads directly supporting the 2nd indel allele
  13. indel line only: # reads supporting a third indel allele

If pileup is invoked without `-c', indel lines and columns between 3 and 7 inclusive will not be outputted.

I see `*' in the pileup sequence column. What are they?

A star at the sequence column represents a deletion. It is a place holder to make sure the number of bases at that column matches the read depth column. Simply ignore `*' if you do not use this information.

I only want to use a subset of alignments in pileup.

If you want to filter on mapping quality, flags, one read group or one library, you may just use the view command. If want to apply more complex filters, you may write an awk command for SAM. For example, I only want to use alignment with two or fewer differences (mismatches+gaps):

samtools view -h aln.bam | perl -ne 'print if (/^@/||(/NM:i:(\d+)/&&$1<=2))' | samtools pileup -S - > out.txt  

or exclude all gapped alignments:

samtools view -h aln.bam | awk '$6!~/[ID]/' | samtools pileup -S -  

Does samtools generate the consensus sequence like Maq?

Yes. Try this:

samtools pileup -cf ref.fa aln.bam | pileup2fq -D100 > cns.fastq  

Again, remember to set -D according to your read depth. Note that pileup2fq applies fewer filters in comparison to varFilter, and you may see tiny inconsistency between the two outputs.

I want to get `unique' alignments from SAM/BAM.

We prefer to say an alignment is `reliable' rather than `unique' as `uniqueness' is not well defined in general cases. You can get reliable alignments by setting a threshold on mapping quality:

samtools view -bq 1 aln.bam > aln-reliable.bam  

You may want to set a more stringent threshold to get more reliable alignments.

In repetitive regions, SAMtools call all bases as 'A' although there are no 'A' bases in reads.

This is due to a floating underflow in the MAQ SNP calling model used by default and only happens in repetitive regions. These calls are always filtered out. However, if you are uncomfortable with this, you may use the simplified SOAPsnp model with:

samtools -avcf ref.fa aln.bam > raw.txt  

The MAQ model and SOAPsnp model usually deliver very similar SNP calls.

How are SNPs and indels called and filtered by SAMtools?

By default, SNPs are called with a Bayesian model identical to the one used in MAQ. A simplified SOAPsnp model is implemented, too. Indels are called with a simple Bayesian model. The caller does local realignment to recover indels that occur at the end of a read but appear to be contiguous mismatches. For an example, see this picture.

The varFilter filters SNPs/indels in the following order:

The first letter indicates the reason why SNPs/indels are filtered when you invoke varFilter with the `-p' option. A SNP/indel filtered by a rule higher in the list will not be tested against other rules.

The Windows version of SAMtools does not work sometimes.

We are sorry that this is due to bugs in the Windows port. The Windows version is mainly meant to be a cross-platform viewer. Most of samtools functionality are not tested. For heavy use of samtools, please run it on Linux machines instead.

For Developers

How to make my aligner work best with samtools?

To get the best performance in SNP calling, we recommend the following rules.

Why mapping quality?

The plot below shows alignment accuracy for 108bp simulated reads under different configurations of BWA. If we only retain `unique' alignment, we get a single spot ungap-se-unqiue which corresponds to ~2300 wrong alignments out of 1.68 million mapped reads. If we look at the mapping quality generated by BWA and set a stringent threshold on that, it is possible to get an accuracy of 400/1.67M (the ungap-se line). That is saying we get >80% fewer false alignments at the cost of 1% loss in sensitivity. Setting a higher threshold further reduces false alignments and helps to reduce noises in identifying structural variations bridging unique regions. The plot may vary with the aligner in use, but it is generally true that an algorithm seeing more suboptimal alignments is more accurate.


Manual Reference Pages - samtools (1)


samtools - Utilities for the Sequence Alignment/Map (SAM) format

bcftools - Utilities for the Binary Call Format (BCF) and VCF


Samtools Commands And Options
Bcftools Commands And Options
Sam Format
Vcf Format
See Also


samtools view -bt ref_list.txt -o aln.bam aln.sam.gz

samtools sort aln.bam aln.sorted

samtools index aln.sorted.bam

samtools idxstats aln.sorted.bam

samtools view aln.sorted.bam chr2:20,100,000-20,200,000

samtools merge out.bam in1.bam in2.bam in3.bam

samtools faidx ref.fasta

samtools pileup -vcf ref.fasta aln.sorted.bam

samtools mpileup -C50 -gf ref.fasta -r chr3:1,000-2,000 in1.bam in2.bam

samtools tview aln.sorted.bam ref.fasta

bcftools index in.bcf

bcftools view in.bcf chr2:100-200 > out.vcf

bcftools view -vc in.bcf > out.vcf 2> out.afs


Samtools is a set of utilities that manipulate alignments in the BAM format. It imports from and exports to the SAM (Sequence Alignment/Map) format, does sorting, merging and indexing, and allows to retrieve reads in any regions swiftly.

Samtools is designed to work on a stream. It regards an input file ‘-’ as the standard input (stdin) and an output file ‘-’ as the standard output (stdout). Several commands can thus be combined with Unix pipes. Samtools always output warning and error messages to the standard error output (stderr).

Samtools is also able to open a BAM (not SAM) file on a remote FTP or HTTP server if the BAM file name starts with ‘ftp://’ or ‘http://’. Samtools checks the current working directory for the index file and will download the index upon absence. Samtools does not retrieve the entire alignment file unless it is asked to do so.


view samtools view [-bchuHS] [-t in.refList] [-o output] [-f reqFlag] [-F skipFlag] [-q minMapQ] [-l library] [-r readGroup] [-R rgFile] <in.bam>|<in.sam> [region1 [...]]

Extract/print all or sub alignments in SAM or BAM format. If no region is specified, all the alignments will be printed; otherwise only alignments overlapping the specified regions will be output. An alignment may be given multiple times if it is overlapping several regions. A region can be presented, for example, in the following format: ‘chr2’ (the whole chr2), ‘chr2:1000000’ (region starting from 1,000,000bp) or ‘chr2:1,000,000-2,000,000’ (region between 1,000,000 and 2,000,000bp including the end points). The coordinate is 1-based.


-b Output in the BAM format.
-f INT Only output alignments with all bits in INT present in the FLAG field. INT can be in hex in the format of /^0x[0-9A-F]+/ [0]
-F INT Skip alignments with bits present in INT [0]
-h Include the header in the output.
-H Output the header only.
-l STR Only output reads in library STR [null]
-o FILE Output file [stdout]
-q INT Skip alignments with MAPQ smaller than INT [0]
-r STR Only output reads in read group STR [null]
-R FILE Output reads in read groups listed in FILE [null]
-S Input is in SAM. If @SQ header lines are absent, the ‘-t’ option is required.
-c Instead of printing the alignments, only count them and print the total number. All filter options, such as ‘-f’, ‘-F’ and ‘-q’ , are taken into account.
-t FILE This file is TAB-delimited. Each line must contain the reference name and the length of the reference, one line for each distinct reference; additional fields are ignored. This file also defines the order of the reference sequences in sorting. If you run ‘samtools faidx <ref.fa>’, the resultant index file <ref.fa>.fai can be used as this <in.ref_list> file.
-u Output uncompressed BAM. This option saves time spent on compression/decomprssion and is thus preferred when the output is piped to another samtools command.

tview samtools tview <in.sorted.bam> [ref.fasta]

Text alignment viewer (based on the ncurses library). In the viewer, press ‘?’ for help and press ‘g’ to check the alignment start from a region in the format like ‘chr10:10,000,000’ or ‘=10,000,000’ when viewing the same reference sequence.

mpileup samtools mpileup [-EBug] [-C capQcoef] [-r reg] [-f in.fa] [-l list] [-M capMapQ] [-Q minBaseQ] [-q minMapQ] in.bam [in2.bam [...]]

Generate BCF or pileup for one or multiple BAM files. Alignment records are grouped by sample identifiers in @RG header lines. If sample identifiers are absent, each input file is regarded as one sample.

In the pileup format (without -uor-g), each line represents a genomic position, consisting of chromosome name, coordinate, reference base, read bases, read qualities and alignment mapping qualities. Information on match, mismatch, indel, strand, mapping quality and start and end of a read are all encoded at the read base column. At this column, a dot stands for a match to the reference base on the forward strand, a comma for a match on the reverse strand, a ’>’ or ’<’ for a reference skip, ‘ACGTN’ for a mismatch on the forward strand and ‘acgtn’ for a mismatch on the reverse strand. A pattern ‘\+[0-9]+[ACGTNacgtn]+’ indicates there is an insertion between this reference position and the next reference position. The length of the insertion is given by the integer in the pattern, followed by the inserted sequence. Similarly, a pattern ‘-[0-9]+[ACGTNacgtn]+’ represents a deletion from the reference. The deleted bases will be presented as ‘*’ in the following lines. Also at the read base column, a symbol ‘^’ marks the start of a read. The ASCII of the character following ‘^’ minus 33 gives the mapping quality. A symbol ‘$’ marks the end of a read segment.

Input Options:

-6 Assume the quality is in the Illumina 1.3+ encoding. -A Do not skip anomalous read pairs in variant calling.
-B Disable probabilistic realignment for the computation of base alignment quality (BAQ). BAQ is the Phred-scaled probability of a read base being misaligned. Applying this option greatly helps to reduce false SNPs caused by misalignments.
-b FILE List of input BAM files, one file per line [null]
-C INT Coefficient for downgrading mapping quality for reads containing excessive mismatches. Given a read with a phred-scaled probability q of being generated from the mapped position, the new mapping quality is about sqrt((INT-q)/INT)*INT. A zero value disables this functionality; if enabled, the recommended value for BWA is 50. [0]
-d INT At a position, read maximally INT reads per input BAM. [250]
-E Extended BAQ computation. This option helps sensitivity especially for MNPs, but may hurt specificity a little bit.
-f FILE The faidx-indexed reference file in the FASTA format. The file can be optionally compressed by razip. [null]
-l FILE BED or position list file containing a list of regions or sites where pileup or BCF should be generated [null]
-q INT Minimum mapping quality for an alignment to be used [0]
-Q INT Minimum base quality for a base to be considered [13]
-r STR Only generate pileup in region STR [all sites]
Output Options:

-D Output per-sample read depth
-g Compute genotype likelihoods and output them in the binary call format (BCF).
-S Output per-sample Phred-scaled strand bias P-value
-u Similar to -g except that the output is uncompressed BCF, which is preferred for piping.

Options for Genotype Likelihood Computation (for -g or -u):

-e INT Phred-scaled gap extension sequencing error probability. Reducing INT leads to longer indels. [20]
-h INT Coefficient for modeling homopolymer errors. Given an l-long homopolymer run, the sequencing error of an indel of size s is modeled as INT*s/l. [100]
-I Do not perform INDEL calling
-L INT Skip INDEL calling if the average per-sample depth is above INT. [250]
-o INT Phred-scaled gap open sequencing error probability. Reducing INT leads to more indel calls. [40]
-P STR Comma dilimited list of platforms (determined by @RG-PL) from which indel candidates are obtained. It is recommended to collect indel candidates from sequencing technologies that have low indel error rate such as ILLUMINA. [all]

reheader samtools reheader <in.header.sam> <in.bam>

Replace the header in in.bam with the header in in.header.sam. This command is much faster than replacing the header with a BAM->SAM->BAM conversion.

cat samtools cat [-h header.sam] [-o out.bam] <in1.bam> <in2.bam> [ ... ]

Concatenate BAMs. The sequence dictionary of each input BAM must be identical, although this command does not check this. This command uses a similar trick to reheader which enables fast BAM concatenation.

sort samtools sort [-no] [-m maxMem] <in.bam> <out.prefix>

Sort alignments by leftmost coordinates. File <out.prefix>.bam will be created. This command may also create temporary files <out.prefix>.%d.bam when the whole alignment cannot be fitted into memory (controlled by option -m).


-o Output the final alignment to the standard output.
-n Sort by read names rather than by chromosomal coordinates
-m INT Approximately the maximum required memory. [500000000]

merge samtools merge [-nur1f] [-h inh.sam] [-R reg] <out.bam> <in1.bam> <in2.bam> [...]

Merge multiple sorted alignments. The header reference lists of all the input BAM files, and the @SQ headers of inh.sam, if any, must all refer to the same set of reference sequences. The header reference list and (unless overridden by -h) ‘@’ headers of in1.bam will be copied to out.bam, and the headers of other files will be ignored.


-1 Use zlib compression level 1 to comrpess the output
-f Force to overwrite the output file if present.
-h FILE Use the lines of FILE as ‘@’ headers to be copied to out.bam, replacing any header lines that would otherwise be copied from in1.bam. (FILE is actually in SAM format, though any alignment records it may contain are ignored.)
-n The input alignments are sorted by read names rather than by chromosomal coordinates
-R STR Merge files in the specified region indicated by STR [null]
-r Attach an RG tag to each alignment. The tag value is inferred from file names.
-u Uncompressed BAM output

index samtools index <aln.bam>

Index sorted alignment for fast random access. Index file <aln.bam>.bai will be created.

idxstats samtools idxstats <aln.bam>

Retrieve and print stats in the index file. The output is TAB delimited with each line consisting of reference sequence name, sequence length, # mapped reads and # unmapped reads.

faidx samtools faidx <ref.fasta> [region1 [...]]

Index reference sequence in the FASTA format or extract subsequence from indexed reference sequence. If no region is specified, faidx will index the file and create <ref.fasta>.fai on the disk. If regions are speficified, the subsequences will be retrieved and printed to stdout in the FASTA format. The input file can be compressed in the RAZF format.

fixmate samtools fixmate <in.nameSrt.bam> <out.bam>

Fill in mate coordinates, ISIZE and mate related flags from a name-sorted alignment.

rmdup samtools rmdup [-sS] <> <out.bam>

Remove potential PCR duplicates: if multiple read pairs have identical external coordinates, only retain the pair with highest mapping quality. In the paired-end mode, this command ONLY works with FR orientation and requires ISIZE is correctly set. It does not work for unpaired reads (e.g. two ends mapped to different chromosomes or orphan reads).


-s Remove duplicate for single-end reads. By default, the command works for paired-end reads only.
-S Treat paired-end reads and single-end reads.

calmd samtools calmd [-EeubSr] [-C capQcoef] <aln.bam> <ref.fasta>

Generate the MD tag. If the MD tag is already present, this command will give a warning if the MD tag generated is different from the existing tag. Output SAM by default.


-A When used jointly with -r this option overwrites the original base quality.
-e Convert a the read base to = if it is identical to the aligned reference base. Indel caller does not support the = bases at the moment.
-u Output uncompressed BAM
-b Output compressed BAM
-S The input is SAM with header lines
-C INT Coefficient to cap mapping quality of poorly mapped reads. See the pileup command for details. [0]
-r Compute the BQ tag (without -A) or cap base quality by BAQ (with -A).
-E Extended BAQ calculation. This option trades specificity for sensitivity, though the effect is minor.

targetcut samtools targetcut [-Q minBaseQ] [-i inPenalty] [-0 em0] [-1 em1] [-2 em2] [-f ref] <in.bam>

This command identifies target regions by examining the continuity of read depth, computes haploid consensus sequences of targets and outputs a SAM with each sequence corresponding to a target. When option -f is in use, BAQ will be applied. This command is only designed for cutting fosmid clones from fosmid pool sequencing [Ref. Kitzman et al. (2010)].

phase samtools phase [-AF] [-k len] [-b prefix] [-q minLOD] [-Q minBaseQ] <in.bam>

Call and phase heterozygous SNPs. OPTIONS:

-A Drop reads with ambiguous phase.
-b STR Prefix of BAM output. When this option is in use, phase-0 reads will be saved in file STR.0.bam and phase-1 reads in STR.1.bam. Phase unknown reads will be randomly allocated to one of the two files. Chimeric reads with switch errors will be saved in STR.chimeric.bam. [null]
-F Do not attempt to fix chimeric reads.
-k INT Maximum length for local phasing. [13]
-q INT Minimum Phred-scaled LOD to call a heterozygote. [40]
-Q INT Minimum base quality to be used in het calling. [13]


view bcftools view [-AbFGNQSucgv] [-D seqDict] [-l listLoci] [-s listSample] [-i gapSNPratio] [-t mutRate] [-p varThres] [-P prior] [-1 nGroup1] [-d minFrac] [-U nPerm] [-X permThres] [-T trioType] in.bcf [region]

Convert between BCF and VCF, call variant candidates and estimate allele frequencies.

Input/Output Options:
Retain all possible alternate alleles at variant sites. By default, the view command discards unlikely alleles.
-b Output in the BCF format. The default is VCF.
-D FILE Sequence dictionary (list of chromosome names) for VCF->BCF conversion [null]
-F Indicate PL is generated by r921 or before (ordering is different).
-G Suppress all individual genotype information.
-l FILE List of sites at which information are outputted [all sites]
-N Skip sites where the REF field is not A/C/G/T
-Q Output the QCALL likelihood format
-s FILE List of samples to use. The first column in the input gives the sample names and the second gives the ploidy, which can only be 1 or 2. When the 2nd column is absent, the sample ploidy is assumed to be 2. In the output, the ordering of samples will be identical to the one in FILE. [null]
-S The input is VCF instead of BCF.
-u Uncompressed BCF output (force -b).
Consensus/Variant Calling Options:
Call variants using Bayesian inference. This option automatically invokes option -e.
-d FLOAT When -v is in use, skip loci where the fraction of samples covered by reads is below FLOAT. [0]
-e Perform max-likelihood inference only, including estimating the site allele frequency, testing Hardy-Weinberg equlibrium and testing associations with LRT.
-g Call per-sample genotypes at variant sites (force -c)
-i FLOAT Ratio of INDEL-to-SNP mutation rate [0.15]
-p FLOAT A site is considered to be a variant if P(ref|D)<FLOAT [0.5]
-P STR Prior or initial allele frequency spectrum. If STR can be full, cond2, flat or the file consisting of error output from a previous variant calling run.
-t FLOAT Scaled muttion rate for variant calling [0.001]
-T STR Enable pair/trio calling. For trio calling, option -s is usually needed to be applied to configure the trio members and their ordering. In the file supplied to the option -s, the first sample must be the child, the second the father and the third the mother. The valid values of STR are ‘pair’, ‘trioauto’, ‘trioxd’ and ‘trioxs’, where ‘pair’ calls differences between two input samples, and ‘trioxd’ (‘trioxs’) specifies that the input is from the X chromosome non-PAR regions and the child is a female (male). [null]
-v Output variant sites only (force -c)
Contrast Calling and Association Test Options:
-1 INT
Number of group-1 samples. This option is used for dividing the samples into two groups for contrast SNP calling or association test. When this option is in use, the following VCF INFO will be outputted: PC2, PCHI2 and QCHI2. [0]
-U INT Number of permutations for association test (effective only with -1) [0]
-X FLOAT Only perform permutations for P(chi^2)<FLOAT (effective only with -U) [0.01]

index bcftools index in.bcf

Index sorted BCF for random access.

cat bcftools cat in1.bcf ["in2.bcf "[..."]]]"

Concatenate BCF files. The input files are required to be sorted and have identical samples appearing in the same order.


Sequence Alignment/Map (SAM) format is TAB-delimited. Apart from the header lines, which are started with the ‘@’ symbol, each alignment line consists of:


Col Field Description
1 QNAME Query template/pair NAME
2 FLAG bitwise FLAG
3 RNAME Reference sequence NAME
4 POS 1-based leftmost POSition/coordinate of clipped sequence
5 MAPQ MAPping Quality (Phred-scaled)
6 CIAGR extended CIGAR string
7 MRNM Mate Reference sequence NaMe (‘=’ if same as RNAME)
8 MPOS 1-based Mate POSistion
9 TLEN inferred Template LENgth (insert size)
10 SEQ query SEQuence on the same strand as the reference
11 QUAL query QUALity (ASCII-33 gives the Phred base quality)
12+ OPT variable OPTional fields in the format TAG:VTYPE:VALUE

Each bit in the FLAG field is defined as:


Flag Chr Description
0x0001 p the read is paired in sequencing
0x0002 P the read is mapped in a proper pair
0x0004 u the query sequence itself is unmapped
0x0008 U the mate is unmapped
0x0010 r strand of the query (1 for reverse)
0x0020 R strand of the mate
0x0040 1 the read is the first read in a pair
0x0080 2 the read is the second read in a pair
0x0100 s the alignment is not primary
0x0200 f the read fails platform/vendor quality checks
0x0400 d the read is either a PCR or an optical duplicate

where the second column gives the string representation of the FLAG field.


The Variant Call Format (VCF) is a TAB-delimited format with each data line consists of the following fields:


Col Field Description
1 CHROM CHROMosome name
2 POS the left-most POSition of the variant
3 ID unique variant IDentifier
4 REF the REFerence allele
5 ALT the ALTernate allele(s), separated by comma
6 QUAL variant/reference QUALity
7 FILTER FILTers applied
8 INFO INFOrmation related to the variant, separated by semi-colon
9 FORMAT FORMAT of the genotype fields, separated by colon (optional)
10+ SAMPLE SAMPLE genotypes and per-sample information (optional)

The following table gives the INFO tags used by samtools and bcftools.


Tag Format Description
AF1 double Max-likelihood estimate of the site allele frequency (AF) of the first ALT allele
DP int Raw read depth (without quality filtering)
DP4 int[4] # high-quality reference forward bases, ref reverse, alternate for and alt rev bases
FQ int Consensus quality. Positive: sample genotypes different; negative: otherwise
MQ int Root-Mean-Square mapping quality of covering reads
PC2 int[2] Phred probability of AF in group1 samples being larger (,smaller) than in group2
PCHI2 double Posterior weighted chi^2 P-value between group1 and group2 samples
PV4 double[4] P-value for strand bias, baseQ bias, mapQ bias and tail distance bias
QCHI2 int Phred-scaled PCHI2
RP int # permutations yielding a smaller PCHI2
CLR int Phred log ratio of genotype likelihoods with and without the trio/pair constraint
UGT string Most probable genotype configuration without the trio constraint
CGT string Most probable configuration with the trio constraint


o Import SAM to BAM when @SQ lines are present in the header:

samtools view -bS aln.sam > aln.bam

If @SQ lines are absent:

samtools faidx ref.fa
samtools view -bt ref.fa.fai aln.sam > aln.bam

where ref.fa.fai is generated automatically by the faidx command.

o Attach the RG tag while merging sorted alignments:

perl -e ’print "@RG\tID:ga\tSM:hs\tLB:ga\tPL:Illumina\n@RG\tID:454\tSM:hs\tLB:454\tPL:454\n"’ > rg.txt
samtools merge -rh rg.txt merged.bam ga.bam 454.bam

The value in a RG tag is determined by the file name the read is coming from. In this example, in the merged.bam, reads from ga.bam will be attached RG:Z:ga, while reads from 454.bam will be attached RG:Z:454.

o Call SNPs and short INDELs for one diploid individual:

samtools mpileup -ugf ref.fa aln.bam | bcftools view -bvcg - > var.raw.bcf
bcftools view var.raw.bcf | varFilter -D 100 > var.flt.vcf

The -D option of varFilter controls the maximum read depth, which should be adjusted to about twice the average read depth. One may consider to add -C50 to mpileup if mapping quality is overestimated for reads containing excessive mismatches. Applying this option usually helps BWA-short but may not other mappers.

o Generate the consensus sequence for one diploid individual:

samtools mpileup -uf ref.fa aln.bam | bcftools view -cg - | vcf2fq > cns.fq

o Call somatic mutations from a pair of samples:

samtools mpileup -DSuf ref.fa aln.bam | bcftools view -bvcgT pair - > var.bcf

In the output INFO field, CLR gives the Phred-log ratio between the likelihood by treating the two samples independently, and the likelihood by requiring the genotype to be identical. This CLR is effectively a score measuring the confidence of somatic calls. The higher the better.

o Call de novo and somatic mutations from a family trio:

samtools mpileup -DSuf ref.fa aln.bam | bcftools view -bvcgT pair -s samples.txt - > var.bcf

File samples.txt should consist of three lines specifying the member and order of samples (in the order of child-father-mother). Similarly, CLR gives the Phred-log likelihood ratio with and without the trio constraint. UGT shows the most likely genotype configuration without the trio constraint, and CGT gives the most likely genotype configuration satisfying the trio constraint.

o Phase one individual:

samtools calmd -AEur aln.bam ref.fa | samtools phase -b prefix - > phase.out

The calmd command is used to reduce false heterozygotes around INDELs.

o Call SNPs and short indels for multiple diploid individuals:

samtools mpileup -P ILLUMINA -ugf ref.fa *.bam | bcftools view -bcvg - > var.raw.bcf
bcftools view var.raw.bcf | varFilter -D 2000 > var.flt.vcf

Individuals are identified from the SM tags in the @RG header lines. Individuals can be pooled in one alignment file; one individual can also be separated into multiple files. The -P option specifies that indel candidates should be collected only from read groups with the @RG-PL tag set to ILLUMINA. Collecting indel candidates from reads sequenced by an indel-prone technology may affect the performance of indel calling.

o Derive the allele frequency spectrum (AFS) on a list of sites from multiple individuals:

samtools mpileup -Igf ref.fa *.bam > all.bcf
bcftools view -bl sites.list all.bcf > sites.bcf
bcftools view -cGP cond2 sites.bcf > /dev/null 2> sites.1.afs
bcftools view -cGP sites.1.afs sites.bcf > /dev/null 2> sites.2.afs
bcftools view -cGP sites.2.afs sites.bcf > /dev/null 2> sites.3.afs

where sites.list contains the list of sites with each line consisting of the reference sequence name and position. The following bcftools commands estimate AFS by EM.

o Dump BAQ applied alignment for other SNP callers:

samtools calmd -bAr aln.bam > aln.baq.bam

It adds and corrects the NM and MD tags at the same time. The calmd command also comes with the -C option, the same as the one in pileup and mpileup. Apply if it helps.


o Unaligned words used in bam_import.c, bam_endian.h, bam.c and bam_aux.c.
o Samtools paired-end rmdup does not work for unpaired reads (e.g. orphan reads or ends mapped to different chromosomes). If this is a concern, please use Picard’s MarkDuplicate which correctly handles these cases, although a little slower.


Heng Li from the Sanger Institute wrote the C version of samtools. Bob Handsaker from the Broad Institute implemented the BGZF library and Jue Ruan from Beijing Genomics Institute wrote the RAZF library. John Marshall and Petr Danecek contribute to the source code and various people from the 1000 Genomes Project have contributed to the SAM format specification.


Samtools website: <>