download_databases

Note

Mirrored from the MetaGEAR Pipeline repository: docs/workflows/download_databases.md (db08c99). Check there directly for up-to-date information.

One-time setup of the reference databases consumed by qc_dna, qc_rna, and microbial_profiles. Run this once per installation (and re-run when you want to refresh a database).

What it does

Downloads and prepares three references, sized at tens of gigabytes each:

• KneadData host reference — the human genome (Bowtie2 index), used for host-read decontamination during QC.
• MetaPhlAn 4 marker database — used for taxonomic profiling.
• HUMAnN 3 ChocoPhlAn (nucleotide, full) and UniRef90 (Diamond-formatted protein) databases — used for functional profiling.

Each tool is downloaded by its native installer wrapped in a Nextflow module, then exported to the destination path you supply in the corresponding *_db parameter.

Inputs

This workflow does not read a samplesheet — there is no --input argument. Instead, you tell it where to put each database via the path parameters below. The metagear-tools CLI manages these paths for you via its config; if you run Nextflow directly you must pass them yourself (typically through a -params-file).

Parameters

Parameter	Type	Default	Controls
--outdir	path	(required)	Output directory for execution metadata (the database files themselves land at the destinations below).
--metaphlan_db	path	—	Destination for the MetaPhlAn 4 marker database.
--humann3_nucleo	path	—	Destination for the HUMAnN 3 ChocoPhlAn nucleotide database.
--humann3_uniref90	path	—	Destination for the HUMAnN 3 UniRef90 Diamond protein database.
--kneaddata_refdb	array	[""]	Destination for the KneadData host reference. Single-element array; the pipeline writes the human-genome Bowtie2 index here.

Output

Each tool writes into the directory you supplied for its *_db parameter. The pipeline does not republish those files into <outdir> (the modules’ publishDir is intentionally suppressed) — the databases stay where the installers placed them.

Typical layout (paths are whatever you passed):

<metaphlan_db>/                      # MetaPhlAn 4 marker DB
<humann3_nucleo>/                    # ChocoPhlAn pangenomes
<humann3_uniref90>/                  # UniRef90 Diamond DB
<kneaddata_refdb[0]>/                # Bowtie2 host genome index

<outdir>/pipeline_info/              # Nextflow execution metadata

Example

nextflow run schirmer-lab/metagear -profile docker \
  --workflow download_databases \
  --outdir databases/ \
  -params-file db-paths.yaml

db-paths.yaml:

metaphlan_db: /data/metagear/metaphlan
humann3_nucleo: /data/metagear/humann/chocophlan
humann3_uniref90: /data/metagear/humann/uniref90_diamond
kneaddata_refdb: [/data/metagear/kneaddata/Homo_sapiens]

Notes

• Disk and time. Total downloads are around 50–70 GB and can take several hours, mostly bound by external download speed.
• GTDB-Tk is not downloaded by this workflow. gene_analysis needs a GTDB-Tk database for its MSP taxonomic assignment step — install it separately and point --gtdb_tk_db at the resulting directory. See the GTDB-Tk installation docs.
• HUMAnN config. The HUMAnN installer is invoked with --update-config no; HUMAnN’s global config is not modified on the host.
• Wrapper convenience. metagear-tools ships templates that wire all of these paths into a single user config — recommended over hand-writing the params file above.