Skip to content

Running MAGmaker on DEMON (WFUSM)

This page covers everything needed to run MAGmaker on the Wake Forest University School of Medicine DEMON cluster. If you have access to DEMON, start here — all databases and environments are already set up.

Prerequisites

  • An active DEMON account with access to the defq partition
  • Access to the Sprockett Lab Isilon share (/isilon/datalake/sprockett_lab/)
  • Git and conda available (both are available cluster-wide)

Getting started

Clone MAGmaker to your project directory on the Isilon share:

cd /isilon/datalake/sprockett_lab/original/WF00SprockettLab/<your_project>/
git clone https://github.com/SprockettLab/MAGmaker.git
cd MAGmaker

Activate the shared Snakemake environment:

conda activate /isilon/datalake/sprockett_lab/original/WF00SprockettLab/envs/envs/snakemake

All pipeline tool environments (FastQC, MEGAHIT, MetaPhlAn, GTDB-tk, etc.) are pre-built and stored in the shared conda prefix — they do not need to be built on first run.

Configuration

The default config.yaml is already configured for DEMON. All databases are pre-downloaded and their paths are set:

Database Path
CheckM2 /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/checkm2/CheckM2_database/uniref100.KO.1.dmnd
GUNC /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/gunc/gunc_db_progenomes2.1.dmnd
GTDB-Tk (release 232) /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/gtdbtk/release232/
MetaPhlAn 4 /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/metaphlan/
Kraken2 Standard /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/kraken2/Kraken2_db_Standard/
geNomad /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/genomad/genomad_db/
CheckV /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/checkv/checkv-db-v1.5/
Human GRCh38 (bowtie2 index) /isilon/datalake/sprockett_lab/original/WF00SprockettLab/dbs/bt2/human_GCA_000001405.29_GRCh38.p14/
Conda environments /isilon/datalake/sprockett_lab/original/WF00SprockettLab/envs/snakemake_envs/

The only things you need to change in config.yaml before running are:

  1. host_filter.genome — path to your host genome FASTA (human GRCh38 bowtie2 index is already built; set db_dir to the path above)
  2. metadata — path to your metadata.txt file

Everything else — CheckM2, GUNC, GTDB-Tk, MetaPhlAn, Kraken2, geNomad, CheckV paths — is already set for DEMON.

See Configuration for the format of these files.

Running the pipeline

The DEMON SLURM profile is at resources/profiles/demon/. It:

  • Submits jobs to the defq partition
  • Uses 8 GB RAM and 120-minute walltime as per-job defaults (overridden for memory-intensive rules)
  • Passes --use-conda and the shared conda prefix automatically
conda activate /isilon/datalake/sprockett_lab/original/WF00SprockettLab/envs/envs/snakemake

# Dry run first — checks the full DAG without submitting jobs
./run_magmaker.sh --profile resources/profiles/demon -n

# Execute
./run_magmaker.sh --profile resources/profiles/demon

Stage by stage

# Stage 1 — Main pipeline
snakemake --profile resources/profiles/demon

# Stage 2 — Generate binning config
snakemake --profile resources/profiles/demon generate_binning_config

# Stage 3 — Binning pipeline through MAG renaming
snakemake --snakefile Snakefile-bin --profile resources/profiles/demon rename_mags \
  --config binning=output/config/auto_binning.txt

# Optional — viral / plasmid track
snakemake --snakefile Snakefile-bin --profile resources/profiles/demon virus_all \
  --config binning=output/config/auto_binning.txt

Monitoring jobs

# Check running jobs
squeue -u $USER

# Watch the Snakemake log (run from the MAGmaker directory)
tail -f logs/$(ls -t logs/ | head -1)

# Cancel all your jobs if needed
scancel -u $USER

SLURM profile defaults

The demon profile (resources/profiles/demon/config.yaml) submits to defq with an 8 GB / 120-minute default, --use-conda with the shared conda prefix, latency-wait: 300, and resources: [kraken_slots=2] (at most two Kraken2 jobs at once — each reads a ~200 GB index).

First-attempt memory overrides it sets (set-resources):

Rule First-attempt memory
host_filter 16 GB
metaspades 256 GB
metaphlan 32 GB
taxonomy_kraken 220 GB
make_concoct_coverage_table, make_metabat2_coverage_table 64 GB (many-BAM references)
run_DAS_Tool, run_gunc, run_genomad 32 GB (DIAMOND / marker search)
run_checkv 16 GB
run_gtdbtk (classify) 128 GB → 256 GB → 320 GB (escalates on retry)

Runtimes are no longer set in the profile — as of 2026-08-19 they moved to runtime_escalate() calls on the rules themselves, which grow the wall-clock request on each retry (see Running the pipeline). Several rules that use DIAMOND also carry retries: in config.yaml because their parallel plumbing fails transiently under high cluster concurrency; a fresh attempt lands in a lower-contention window and clears it.

To further override resources for a single rule:

snakemake --profile resources/profiles/demon \
  --set-resources megahit:mem_mb=512000 megahit:runtime=2880

defq partition limits

  • Max walltime: 90 days
  • Default memory: 4 GB/CPU (overridden explicitly by the profile)
  • Snakemake submits up to 100 jobs simultaneously

Private analysis files

Project-specific metadata.txt and binning.txt files should be named with a project prefix (e.g., MyProject_metadata.txt) and stored in resources/config/. The .gitignore excludes *_metadata.txt, *_binning.txt and *_config.yaml to prevent accidental commits of data containing patient or project identifiers.

Logs

  • Snakemake workflow logs: logs/<timestamp>.snakemake.log
  • Rule-level tool logs: output/logs/{rule}/{sample}.log
  • SLURM job logs: .snakemake/slurm_logs/{rule}/{sample}/

← Documentation home