Installation

Requirements

• Linux or macOS (the binning stage's CONCOCT env is Linux-first; see the note below)
• conda or mamba
• ~30–40 GB disk space for conda environments (built automatically on first run)
• A host genome FASTA for host read removal (e.g., human GRCh38, mouse GRCm39)
• Reference databases are separate and much larger — GTDB-Tk ~100 GB, Kraken2 Standard ~200 GB, geNomad ~1.5 GB, CheckV ~1.5 GB, CheckM2 / GUNC / MetaPhlAn a few GB each. See Database setup.

All tool dependencies (FastQC, fastp, MEGAHIT, MetaPhlAn 4, sourmash, MetaBAT2, MaxBin2, CONCOCT, SemiBin2, DAS_Tool, Binette, CheckM2, GUNC, GTDB-Tk, CMSeq, geNomad, CheckV, …) are installed automatically by Snakemake into isolated conda environments on first run.

---

Setup

# 1. Clone the repository
git clone https://github.com/SprockettLab/MAGmaker.git
cd MAGmaker

# 2. Install mamba if you don't already have it
conda install -n base -c conda-forge mamba

# 3. Create and activate the Snakemake environment
mamba env create -n snakemake -f resources/env/snakemake.yaml
conda activate snakemake

All other environments (qc, assemble, profile, binning, mag_qc, etc.) are created automatically the first time a rule that needs them runs. Pass --use-conda when running locally; the demon profile handles this automatically.

---

Test data

The repository includes a minimal test dataset to verify the pipeline is installed and configured correctly:

• Reads: resources/test/test_reads/{John,Paul,George,Ringo}_{R1,R2}.fastq.gz — tiny FASTQ files
• Host index: resources/test/test_dbs/GCA_000001635.9.* — pre-built bowtie2 index for mouse chromosome 1 (tiny subset, sufficient for testing host filter logic)
• Single-end / mixed-layout examples: resources/test/metadata_single_end.txt and resources/test/metadata_mixed_layout.txt run against the same test reads; python3 resources/test/test_metadata_layout.py exercises the metadata loader.

The default metadata.txt and binning.txt point to this test data. To run a test:

conda activate snakemake

# Dry run first
snakemake --cores 4 --use-conda -n

# Execute
snakemake --cores 4 --use-conda

To use the test host index, set in config.yaml:

host_filter:
  db_dir: resources/test/test_dbs
  genome: resources/db/bt2/GCA_000001635.9.fna

---

Conda environments

Each pipeline module has its own environment YAML in resources/env/:

File	Used by
snakemake.yaml	top-level (install this first)
fastp.yaml	qc.smk — default trimmer (paired + single-end)
cutadapt.yaml	qc.smk — non-default trimmer
fastqc.yaml	qc.smk — all FastQC rules
qc.yaml	qc.smk — host filter and index build (bowtie2, samtools, pigz)
assemble.yaml	assemble.smk — MEGAHIT, metaSPAdes, QUAST
mapping.yaml	mapping.smk, and cmseq.smk's mapping step — minimap2, bowtie2, samtools
prototype_selection.yaml	prototype_selection.smk — sourmash, scikit-bio
profile.yaml	profile.smk (MetaPhlAn 4, Kraken2, Bracken, Krona) and cmseq.smk's poly.py
binning.yaml	binning.smk — MetaBAT2, MaxBin2, FragGeneScan
concoct_linux.yaml	binning.smk — CONCOCT (Linux); on macOS point the conda: at concoct_osx.yaml
semibin.yaml	binning.smk — SemiBin2
selected_bins.yaml	selected_bins.smk — DAS_Tool, Fasta_to_Contig2Bin
binette.yaml	selected_bins.smk — Binette (default consolidation tool)
checkm2.yaml	mag_qc.smk — CheckM2 (separate env; Python version gap)
gunc.yaml	mag_qc.smk — GUNC (separate env; conflicts with CheckM2)
gtdbtk.yaml	mag_qc.smk — GTDB-Tk (separate env; dependency conflicts)
mag_qc.yaml	mag_qc.smk, cmseq.smk, virus.smk summary/aggregation scripts (pandas)
genomad.yaml	virus.smk — geNomad
checkv.yaml	virus.smk — CheckV

resources/env/ also holds a few unwired alternates and single-tool variants (concoct_osx.yaml, bowtie2.yaml, metaphlan.yaml, sourmash.yaml, binning_concoct_*.yaml) that no rule references by default.

Note: tbb=2020.2 is pinned in qc.yaml and mapping.yaml because newer TBB versions break bowtie2 on some systems. Do not remove this pin without testing.

---

← Documentation home