The pbwt package provides a core implementation and development environment for PBWT (Positional Burrows-Wheeler Transform) methods for storing and computing on genome variation data sets.
More precisely, PBWT supports a run-length compressed representation of aligned haplotype data, on which efficient matching algorithms can be built. Typically PBWT compression is much better than generic compression, particularly for large numbers of haplotypes, and search algorithms are linear in the query size independent of reference size.
A description of the basic data structure and matching algorithms is given in "Efficient haplotype matching and storage using the Positional Burrows-Wheeler Transform (PBWT)", Richard Durbin Bioinformatics 30:1266-72 (2014).
There are various phasing and imputation methods in the software that are not yet published.
Richard Durbin [email protected]
May 2013, updated September 2014
April 2024: the -paint and -paintSparse commands are described in Yaoling Yang, Richard Durbin, Astrid K. N. Iversen, and Daniel J. Lawson. 2024. “Sparse Haplotype-Based Fine-Scale Local Ancestry Inference at Scale Reveals Recent Selection on Immune Responses.” medRxiv 2024.03.13.24304206
Download htslib from https://github.com/samtools/htslib, and compile it
git clone https://github.com/samtools/htslib cd htslib make cd ..
Download and make pbwt
git clone https://github.com/richarddurbin/pbwt cd pbwt make
Typing
pbwt
by itself gives a list of commands with brief descriptions.
A quick synopsis for usage is:
macs 11000 1e6 -t 0.001 -r 0.001 > 11k.macs
pbwt -checkpoint 10000 -readMacs 11k.macs -write macs11k.pbwt -writeSites macs.sites
NB "checkpoint 10000" writes out files every 10000 sites during the vcfq conversion to alternating checkA.{pbwt,sites} and checkB.{pbwt,sites} files.
pbwt -read macs11k.pbwt -subsample 0 10000 -write macs10k.pbwt
pbwt -read macs11k.pbwt -subsample 10000 1000 -write macs1k.pbwt
pbwt -read macs10k.pbwt -sfs > macs10k.sfs
gives the site frequency spectrum for macs10k
pbwt -read macs1k.pbwt -haps macs1k.haps
writes out the haplotypes stored in macs1k
pbwt -read macs10k.pbwt -matchDynamic macs1k.pbwt > macs1k-10k.max
for each sequence in macs1k, finds maximal matches to anything in macs10k
pbwt -read macs10k.pbwt -maxWithin > macs10k.max
finds maximal matches for each sequence in macs10k to anything else in macs10k
To start from real data in a .vcf file rather than a macs simulation use
pbwt -checkpoint 10000 -readVcfGT data.vcf -writeAll data
Note that -writeAll xxx will write xxx.pbwt, xxx.sites, xxx.samples and any other associated files, and -readAll xxx will correspondingly read xxx.pbwt and any available files based on suffix.
pbwt is very happy to handle up to 100,000 haplotypes, probably a million.