bwa

[!Note] BWA-MEM has been replaced by minimap2 for long reads and will be replaced by minibwa for short reads. If you still prefer the exact legacy output of BWA-MEM, try BWA-MEM2 or the compatibility mode of BWA-MEM3.

Getting started

git clone https://github.com/lh3/bwa.git
cd bwa; make
./bwa index ref.fa
./bwa mem ref.fa read-se.fq.gz | gzip -3 > aln-se.sam.gz
./bwa mem ref.fa read1.fq read2.fq | gzip -3 > aln-pe.sam.gz

Introduction

BWA is a software package for mapping DNA sequences against a large reference genome, such as the human genome. It consists of three algorithms: BWA-backtrack, BWA-SW and BWA-MEM. The first algorithm is designed for Illumina sequence reads up to 100bp, while the rest two for longer sequences ranged from 70bp to a few megabases. BWA-MEM and BWA-SW share similar features such as the support of long reads and chimeric alignment, but BWA-MEM, which is the latest, is generally recommended as it is faster and more accurate. BWA-MEM also has better performance than BWA-backtrack for 70-100bp Illumina reads.

For all the algorithms, BWA first needs to construct the FM-index for the reference genome (the index command). Alignment algorithms are invoked with different sub-commands: aln/samse/sampe for BWA-backtrack, bwasw for BWA-SW and mem for the BWA-MEM algorithm.

Availability

BWA is released under GPLv3. The latest source code is freely available at github. Released packages can be downloaded at SourceForge. After you acquire the source code, simply use make to compile and copy the single executable bwa to the destination you want. The only dependency required to build BWA is zlib.

Since 0.7.11, precompiled binary for x86_64-linux is available in bwakit. In addition to BWA, this self-consistent package also comes with bwa-associated and 3rd-party tools for proper BAM-to-FASTQ conversion, mapping to ALT contigs, adapter triming, duplicate marking, HLA typing and associated data files.

Seeking help

The detailed usage is described in the man page available together with the source code. You can use man ./bwa.1 to view the man page in a terminal. The HTML version of the man page can be found at the BWA website. If you have questions about BWA, you may sign up the mailing list and then send the questions to [email protected]. You may also ask questions in forums such as BioStar and SEQanswers.

Citing BWA

• Li H. and Durbin R. (2009) Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics, 25, 1754-1760. [PMID: 19451168]. (if you use the BWA-backtrack algorithm)

• Li H. and Durbin R. (2010) Fast and accurate long-read alignment with Burrows-Wheeler transform. Bioinformatics, 26, 589-595. [PMID: 20080505]. (if you use the BWA-SW algorithm)

• Li H. (2013) Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv:1303.3997v2 [q-bio.GN]. (if you use the BWA-MEM algorithm or the fastmap command, or want to cite the whole BWA package)

Please note that the last reference is a preprint hosted at arXiv.org. I do not have plan to submit it to a peer-reviewed journal in the near future.

Frequently asked questions (FAQs)

1. What types of data does BWA work with?
2. Why does a read appear multiple times in the output SAM?
3. Does BWA work on reference sequences longer than 4GB in total?
4. Why can one read in a pair has high mapping quality but the other has zero?
5. How can a BWA-backtrack alignment stands out of the end of a chromosome?
6. Does BWA work with ALT contigs in the GRCh38 release?
7. Can I just run BWA-MEM against GRCh38+ALT without post-processing?
8. Why does BWA use a lot of memory?

1. What types of data does BWA work with?

BWA works with a variety types of DNA sequence data, though the optimal algorithm and setting may vary. The following list gives the recommended settings:

• Illumina/454/IonTorrent single-end reads longer than ~70bp or assembly contigs up to a few megabases mapped to a closely related reference genome:

    bwa mem ref.fa reads.fq > aln.sam
  

• Illumina single-end reads shorter than ~70bp:

    bwa aln ref.fa reads.fq > reads.sai; bwa samse ref.fa reads.sai reads.fq > aln-se.sam
  

• Illumina/454/IonTorrent paired-end reads longer than ~70bp:

    bwa mem ref.fa read1.fq read2.fq > aln-pe.sam
  

• Illumina paired-end reads shorter than ~70bp:

    bwa aln ref.fa read1.fq > read1.sai; bwa aln ref.fa read2.fq > read2.sai
    bwa sampe ref.fa read1.sai read2.sai read1.fq read2.fq > aln-pe.sam
  

• PacBio subreads or Oxford Nanopore reads to a reference genome:

    bwa mem -x pacbio ref.fa reads.fq > aln.sam
    bwa mem -x ont2d ref.fa reads.fq > aln.sam
  

BWA-MEM is recommended for query sequences longer than ~70bp for a variety of error rates (or sequence divergence). Generally, BWA-MEM is more tolerant with errors given longer query sequences as the chance of missing all seeds is small. As is shown above, with non-default settings, BWA-MEM works with Oxford Nanopore reads with a sequencing error rate over 20%.

2. Why does a read appear multiple times in the output SAM?

BWA-SW and BWA-MEM perform local alignments. If there is a translocation, a gene fusion or a long deletion, a read bridging the break point may have two hits, occupying two lines in the SAM output. With the default setting of BWA-MEM, one and only one line is primary and is soft clipped; other lines are tagged with 0x800 SAM flag (supplementary alignment) and are hard clipped.

3. Does BWA work on reference sequences longer than 4GB in total?

Yes. Since 0.6.x, all BWA algorithms work with a genome with total length over 4GB. However, individual chromosome should not be longer than 2GB.

4. Why can one read in a pair have a high mapping quality but the other has zero?

This is correct. Mapping quality is assigned for individual read, not for a read pair. It is possible that one read can be mapped unambiguously, but its mate falls in a tandem repeat and thus its accurate position cannot be determined.

5. How can a BWA-backtrack alignment stand out of the end of a chromosome?

Internally BWA concatenates all reference sequences into one long sequence. A read may be mapped to the junction of two adjacent reference sequences. In this case, BWA-backtrack will flag the read as unmapped (0x4), but you will see position, CIGAR and all the tags. A similar issue may occur to BWA-SW alignment as well. BWA-MEM does not have this problem.

6. Does BWA work with ALT contigs in the GRCh38 release?

Yes, since 0.7.11, BWA-MEM officially supports mapping to GRCh38+ALT. BWA-backtrack and BWA-SW don’t properly support ALT mapping as of now. Please see README-alt.md for details. Briefly, it is recommended to use bwakit, the binary release of BWA, for generating the reference genome and for mapping.

7. Can I just run BWA-MEM against GRCh38+ALT without post-processing?

If you are not interested in hits to ALT contigs, it is okay to run BWA-MEM without post-processing. The alignments produced this way are very close to alignments against GRCh38 without ALT contigs. Nonetheless, applying post-processing helps to reduce false mappings caused by reads from the diverged part of ALT contigs and also enables HLA typing. It is recommended to run the post-processing script.

8. Why does BWA use a lot of memory?

This is typically caused by FASTQ generated from a coordinate-sorted BAM. BWA uses a lot more memory for centromeric reads than for unique reads. In a FASTQ file generated from a sequencing run, centromeric reads are rare in each batch and rarely cause troubles. However, in a coordinate-sorted FASTQ file, a whole batch could consist of centromeric reads. Such a batch will take a lot more memory and time to map; the insert size estimate will be distorted as well. General rule: NEVER use Picard SamToFastq on coordiate-sorted BAM; use samtools collate+fastq instead.