Installation

Installation

HashBrown is a Python package (≥ 3.10) that drives command-line bioinformatics programs. The package needs only psutil; the programs come from bioconda.

From bioconda (recommended)

conda create -n hashbrown -c conda-forge -c bioconda hashbrown
conda activate hashbrown

hashbrown --version

The bioconda package installs HashBrown together with every core program the three pipelines use; nothing else is needed for assembly, polishing and the base report.

From source

git clone https://github.com/duceppemo/HashBrown
cd HashBrown
conda env create -f environment.yml    # or: mamba env create -f environment.yml
conda activate HashBrown
pip install .

hashbrown --version

environment.yml installs every program the three pipelines use. If you only ever assemble short reads, you can trim it down to skesa/spades, fastp, samtools, minimap2, quast, mosdepth and bandage.

From PyPI

The package itself is on PyPI as hashbrown-assembly, so once an environment with the programs is active you can skip the clone:

pip install hashbrown-assembly

Without installing

From a clone with the environment active, python hashbrown.py … behaves exactly like the hashbrown command.

Optional programs

• Bandage draws PNG pictures of the assembly graphs.
• gfa_connector (shipped with the bioconda SKESA package) rebuilds an assembly graph for SKESA assemblies, which have none natively.
• The report's deeper QC modules (pycoQC-ng, dnaapler, Merqury, species ID, CheckM2, BUSCO, MLST, MOB-suite, AMRFinderPlus, abricate, Bakta/PGAP, ConFindr, GenomeScope2) are each optional — install one and it runs. Install one-liners and the hashbrown db database downloads are listed under Usage → Optional QC modules.

If any is missing, HashBrown says so once and skips that output; nothing else changes.

Troubleshooting

• The conda solver takes forever or fails. Use mamba, and make sure the channels are exactly conda-forge then bioconda (as in environment.yml). Historically Medaka's TensorFlow builds were the culprit — the reason BGA and BGA_short were separate tools — but Medaka ≥ 2.0 is PyTorch-based and coexists with SKESA/SPAdes.
• Some required programs are not installed or not on the PATH. Activate the environment (conda activate HashBrown). The message lists exactly which programs are missing; HashBrown only requires the ones your flags use.
• Medaka polishing never finishes or crashes. Check that bcftools and samtools load (samtools version): a broken conda install of those is the usual cause. Reinstalling the environment fixes it.