Aligns reads to a Variation Graph by combining Seed-and-Extend approaches with Partial Order Alignment (POA).
There are three main steps in this program:
- Indexing - We first build an index over the input graph, which contains all of the kmers (of a given size k) encoded by it, and their positions w.r.t. the linearized forward/reverse sequence. These positions are stored in a memory-efficient Minimal Perfect Hash Function.
- Mapping - We then split the query into kmers, and find perfect maches between these kmers and the linearization ("anchors"). Anchors that are close together are grouped into approximate alignments ("chains"). This step is inspired by Minimap2.
- Alignment - In order to convert chains into base-level alignments, we align the subgraph implied by the chain to the query. We use abPOA (or more specifically its Rust bindings) to perform the sequence-to-graph alignment.
This crate is mostly inteded to be used as an executable, but you can also use its functions if you may want to do so.
In order to install vgaligner as an executable, clone this repository, move to its main folder and run:
cargo install --path .
Graphs must be sorted in order to ensure that vgaligner works correctly. You can do so with odgi:
odgi sort -i unsorted_graph.gfa -o - -p Ygs -P | odgi view -i - -g > sorted_graph.gfa
Generate the index with the following command:
vgaligner index -k {kmer-size} -i {input.gfa} -o {index.idx}
There are also additional parameters to limit the number of nodes/edges to be traversed, you can use
vgaligner index --help to see them all.
Map reads to the graph with the following command:
vgaligner map -i {index.idx} -f {reads.fa/fq} -o {output.gaf}
This will output the chains in GAF format.
If you also want to perform
the alignment with abPOA (this is highly recommended!), you will need to pass the --also-align parameter.
This will generate an additional GAF file which contains the alignment.
You can see all of the available parameters with vgaligner map --help.
This tool has been tested on graphs obtained from HLA-zoo.
The validation pipeline uses snakemake, and can be executed by using the following commands:
cd experiments-snakemake && snakemake -k --cores {cores_to_be_used}
We recommend having at least 16 GB RAM and a CPU with 4+ cores.