hashbrown-assembly 1.1.1

HashBrown assembles bacterial genomes from whatever reads you have: Nanopore long reads, Illumina (or Ion Torrent) short reads, or both. With long reads it assembles with Flye or Shasta and polishes with Medaka; add short reads and they polish the assembly further (Polypolish, then pyPOLCA --careful); with short reads alone it assembles with SKESA or SPAdes. Every run ends with QC — QUAST metrics, mapping QC (samtools flagstat), read depth (mosdepth), assembly graphs, before/after polishing dot plots — and a per-sample HTML report (with a PDF twin) showing the effect of every stage on the reads and the assembly, plus optional modules for consensus QV (Merqury), species confirmation (mash + ANI), completeness (CheckM2, BUSCO), typing (MLST, MOB-suite), AMR and virulence genes (AMRFinderPlus, abricate), annotation (Bakta or PGAP) and contamination (ConFindr, GenomeScope2, blob plot) — each one auto-detected and skipped with a hint when not installed.

Like its namesake — shredded potatoes pressed into one piece — HashBrown presses reads into one assembly. It merges and supersedes BGA and BGA_short, which were split when Medaka's TensorFlow builds conflicted with the short-read tools; one conda environment now covers everything.

Quick start

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

# Long reads only
hashbrown -l nanopore_fastq/ -o results/

# Hybrid: the short reads polish the long-read assemblies
hashbrown -l nanopore_fastq/ -s illumina_fastq/ -o results/

# Short reads only
hashbrown -s illumina_fastq/ -o results/ -a skesa

The bioconda package brings the whole core toolchain with it. The Python package alone is on PyPI as hashbrown-assembly (pip install hashbrown-assembly), and a source install (git clone + conda env create -f environment.yml + pip install .) works too — see the wiki's Installation page.

Each long-read fastq file is one sample. Short-read files pair on _R1/_R2 in the name; a file without either is a single-end sample. The sample name is everything before the first underscore, and in hybrid mode the long- and short-read names must match. Rerunning the same command resumes where it stopped.

Documentation

Everything else is in the wiki, whose sources are maintained in docs/wiki:

• Installation — conda, pip, troubleshooting
• Usage — inputs, options, resuming, performance
• Methods — what each step runs and why
• Outputs — folders and files, with example reports from a real E. coli K-12 hybrid run (HTML · PDF) and from a deliberately contaminated sample (HTML · PDF)
• Validation — benchmarking the polisher combinations against a gold standard
• FAQ — troubleshooting
• Development — tests, CI, releases

Citation

If HashBrown is useful in your work, please cite this repository (see CITATION.cff) and the programs it runs: Flye or Shasta, Medaka, Porechop, Filtlong, SKESA or SPAdes, fastp, Polypolish, pyPOLCA, BWA, minimap2, SAMtools, QUAST, mosdepth, MUMmer and Bandage (references).

License

MIT