pku-embl/metagenomics-skills/binchicken

Bin Chicken

Upstream: AroneyS/binchicken

Citation

Aroney, S. T. N., Newell, R. J. P., Tyson, G. W. & Woodcroft, B. J. Bin Chicken: targeted metagenomic coassembly for the efficient recovery of novel genomes. Nat. Methods (2025). https://doi.org/10.1038/s41592-025-02901-1

See also docs/references.md.

Analytical thinking

Low-abundance genomes are often below recoverable coverage in any single sample. Blind coassembly of many samples increases graph complexity and compute; random subsets explode combinatorially.

Bin Chicken plans targeted coassemblies: match SingleM marker sequences across metagenomes, optionally prioritize markers divergent from references, then greedily cluster samples into coassembly groups. It is a sample-selection / assembly-planning tool — not a replacement for MetaBAT2/BASALT after contigs exist.

Use when:

• Mining public repositories or large private cohorts for rare/novel lineages
• You need a principled alternative to metadata-only pooling
• You want novelty-weighted recovery (markers distant from known genomes)

Do not use Bin Chicken to justify pooling unrelated biological strata for a single hypothesis-driven contrast study — see study-design.

Prerequisites

How to run

Create forward and reverse lists with one matching absolute FASTQ path per line, in identical sample order. Bin Chicken currently targets Illumina paired reads.

set -euo pipefail
[[ -s samples_forward.txt && -s samples_reverse.txt ]] || exit 2
[[ $(wc -l < samples_forward.txt) -eq $(wc -l < samples_reverse.txt) ]] || exit 3
binchicken --version
binchicken single \
  --forward-list samples_forward.txt --reverse-list samples_reverse.txt \
  --run-aviary --cores 64 --output binchicken_single_assembly \
  >binchicken.single.stdout.log 2>binchicken.single.stderr.log
binchicken iterate \
  --coassemble-output binchicken_single_assembly \
  --run-aviary --cores 64 --output binchicken_2_coassembly \
  >binchicken.iterate.stdout.log 2>binchicken.iterate.stderr.log

A later iteration can set --coassembly-samples 3 and use the preceding iteration as --coassemble-output. Do not launch the expensive Aviary path before reviewing the dry plan, storage, and scheduler resources.

Decision tree

Low-abundance / novelty MAG recovery across many samples?
├─ Need automated coassembly targeting → binchicken
├─ Few related samples, clear strata → study-design + matched co-assembly
├─ After contigs exist → basalt ★ / metawrap / dastool / single binners
└─ Only abundance of known refs → coverm / sylph / metaphlan (not coassembly)

Methods caveats

• Targeted coassembly can still create chimeras — multi-coverage binning and

GUNC/CheckM2 remain mandatory.

• Report marker DB, novelty filter, group size, and assembler.
• Genome counts from large public runs are not transferable to your biome.

Validation, recovery, and provenance

Require all input libraries to be accounted for, emitted coassembly groups to contain only permitted strata, SingleM evidence for targets, completed Aviary logs, and explicit recovered/failed group status. Compare individual and targeted coassemblies under identical assembly, mapping, CheckM2/GUNC and dereplication gates; report recovery gains, chimerism and compute—not raw MAG count alone.

Resume only through documented Bin Chicken/Aviary checkpoints with unchanged lists, databases and parameters; otherwise use a fresh output. Record Bin Chicken/SingleM/Aviary versions, metapackage/GTDB release, ordered read lists and checksums, commands, group manifests, seeds, resource use and accepted/rejected MAG lineage. Targeted grouping does not prove that coassembled contigs belong to one population.

Related skills

singlem · study-design · megahit · metaspades · basalt · coverm · gtdbtk · tool-selection · metagenomics-workflow