By David Cook Wildlife Photography - originally posted to Flickr as Galah (Eolophus roseicapillus), CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=8388233
Galah aims to be a more scalable metagenome assembled genome (MAG) dereplication method. That is, it clusters microbial genomes together based on their average nucleotide identity (ANI), and chooses a single member of each cluster as the representative.
Galah uses a greedy clustering approach to speed up genome dereplication, relative to e.g. dRep, particularly when there are many closely related genomes (i.e. >95% ANI). Generated cluster representatives have 2 properties. If the ANI threshold was set to 99%, then:
- Each representative is <99% ANI to each other representative.
- All members are >=99% ANI to the representative.
If CheckM genome qualities were specified, then the clusters have an additional property:
- Each representative genome has a better quality score than other members of the cluster. Each genome is assigned a quality score based on the formula "completeness - 4*contamination".
If CheckM qualities are not used, then:
- Each representative genome was specified to galah before other members of the cluster.
The overall greedy clustering approach was largely inspired by the work of Donovan Parks, as described in this publication. It operates in 3 steps. In the first step, genomes are assigned as representative if no genomes of higher quality are >99% ANI. In the second step, each non-representative genome is assigned to the representative genome it has the highest ANI with.
CoverM can be installed through the bioconda conda channel. After initial setup of conda and the bioconda channel, it can be installed with
conda install galah
Galah can also be used indirectly through CoverM, which is also available on bioconda.
Galah can be installed by downloading statically compiled binaries, available on the releases page.
Third party dependencies listed below are required for this method.
Galah can also be installed from source, using the cargo build system after installing Rust.
cargo install galah
Third party dependencies listed below are required for this method.
To run an unreleased version of Galah, after installing Rust:
git clone https://github.com/wwood/galah
cd galah
cargo run -- cluster ...etc...
Third party dependencies listed below are required for this method.
Galah relies on these 3rd party tools, which must be installed separately.
- Dashing v0.4.0 https://github.com/dnbaker/dashing
- FastANI v1.3 https://github.com/ParBLiSS/FastANI
For clustering a set of genomes at 99% ANI:
galah cluster --genome-fasta-files /path/to/genome1.fna /path/to/genome2.fna >clusters
There are several other options for specifying genomes, ANI cutoffs, etc. See
galah cluster --help for more information.
Similar to dRep, galah operates in two stages. In the first, a fast pre-clustering distance (dashing) is calculated between each pair of genomes. Genome pairs are only considered as potentially in the same cluster with FastANI if the prethreshold ANI is greater than the specified value. By default, the precluster ANI is set at 95% and the final ANI is set at 99%.
Galah is made available under GPL3+. See LICENSE.txt for details. Copyright Ben Woodcroft.
Developed by Ben Woodcroft at the Australian Centre for Ecogenomics.