weaver

Source code is available upon request and will be made publicly available prior to publication

Weaver is a short read mapper for pangenome references. Weaver supports both linear (FASTA) and graphs (rGFA) inputs and additionally uses small variants in a supplementary VCF file. It is designed to reduce reference bias in variant calling while being a simple drop-in replacement for linear reference mappers.

Getting started

Installation

Go to the release page, download the latest statically linked x86_64 binary and add executable permissions. The binary should run on any 64-bit Linux OS. It does not require a GPU, FPGA or any non-standard hardware. Binary was built and tested on Red Hat Enterprise 9 (RHEL9).

mkdir -p bin && cd bin
wget https://github.com/DecodeGenetics/weaver/releases/download/v0.2.0/weaver
chmod a+x weaver

Optionally you may consider adding the new bin directory to your $PATH.

Usage

Weaver has two main subcommands, weaver index to construct the Weaver index and weaver map to map the reads to the pangenome. The Weaver index is constructed once and loaded during mapping. Typical usage:

# Builds an weaver minimizer index (wmi) once at genome.gfa.gz.wmi
weaver index genome.gfa.gz --vcf=small_variants.vcf.gz

# Maps paired short-reads to the graph
weaver map genome.gfa.gz interleaved.fq.gz > out.sam

# Output contains all necessary SAM tags for samtools markdup so it can be piped directly through samtools into a BAM/CRAM (requires samtools)
weaver map genome.gfa.gz interleaved.fq.gz | samtools markdup -u - - | samtools view --remove-tag ms -b -o final.bam
samtools index final.bam # Output is already position sorted

# The alignments are in context of the stable sequences of the graph, a linear representation of the GFA. You can make a FASTA file with those sequences with (requires gfatools)
gfatools gfa2fa -s genome.gfa.gz > genome.fa
samtools faidx genome.fa

Build from source code

Requirements: Linux OS (64bit), C++17 compiler (GCC 7+ or clang 12+), CMake 3.2+, libzstd, libz. Tested on RHEL9 OS (64bit).

Recursively git clone the repo and then build with:

git clone --recursive https://github.com/DecodeGenetics/weaver && cd weaver
mkdir -p build && cd build
cmake .. # Builds a release build by default
make -j4 weaver # Go grab some coffee, building takes a couple of minutes...

I recommend to retrying in an empty build directory if you encounter any errors from cmake, i.e. some dependency not found.

Demo

If you want to run a demo with your prebuilt binary or build, you can run the following (working directory should still be your bin or build directory)

$ ./weaver version # Check version
0.2.0-5aabbfa
5aabbfac6d86953c78e15e79a36bfcc28d5fa784
$ ./weaver index ../test/data/test_human_10k.gfa.gz --threads 4 -k 17 -w 5 --vcf=../test/data/truth.vcf.gz --log=./demo_weaver_index.log --vverbose
$ ./weaver idxstats ../test/data/test_human_10k.gfa.gz.wmi # Print some index stats
Index stats:
  k = 17
  w = 5
  graph size = 3363 bytes
  map size = 3512 keys
$ ./weaver map ../test/data/test_human_10k.gfa.gz ../test/data/small.read1.fq.gz --fq2=../test/data/small.read2.fq.gz \
    --extra-header-lines=../test/data/extra_header_lines.tsv --threads=4 > ./output.sam
$ grep -v ^@PG ./output.sam | md5sum  # Requires grep and md5sum
7bf29e8b638070e95c9a1313c1eb6f53  -

Use the default values for k and w if when running on the full human genome. Details about the available options are in ./weaver [subcommand] --help

Unit tests

To run the unit test suite use (possible when source code has been released):

make # Compiles everything, including the units test
make test # Runs the unit tests

License

MIT License