/ Nextflow · nf-core template · v0.1.0-dev
nf-core-soillifeatlas takes raw LC-MS lipidomics spectra and returns a quantitative readout of bacteria, fungi, plants, archaea, animals and protozoa — a cross-kingdom source-decomposition pipeline in a single command.
§ 02 · Workflow
The pipeline implements the APPLY path for cross-kingdom source decomposition, with an IS/RIE/ref-filter correction stack on top. Every node below maps one-to-one onto a Nextflow module, which wraps one function in framework/*.py.
stage 01
MS2 · feature table
A consensus feature table (Parquet) and MS2 spectra (MGF) produced upstream by MZmine 4.9.14 IIMN + GNPS2 FBMN. Optional sample metadata (TSV) unlocks treatment-effect verification.
inputs
stage 02
SIMPER fingerprint atlas
Sample spectra are matched against the v0.1 SIMPER fingerprint atlas — 12,710 cross-batch features across 18 phyla — using matchms cosine ≥0.7, precursor ≤5 ppm, ≥4 matched peaks. Decomposition depends only on feature_id matching, not lipid annotations.
artefacts
stage 03
IS · RIE(floor=0.20) · ref-filter
Three correction layers from the analysis-19 feedback loop. L2 scales intensities by the spiked internal standard (default LPE 18:0(d7), 100 pmol). L3 applies the response-ionization-efficiency correction with a floor of 0.20 to block over-amplification of weak ionizers. L5 restricts archaea to the ArchLips reference set.
artefacts
stage 04
4 methods in parallel
The corrected matrix is decomposed against the SIMPER atlas using four methods run in parallel: non-negative least squares, standard bray-curtis, enriched-weighted bray-curtis, and fold-change-weighted bray-curtis (the primary method). Methods are checkable — compare results side-by-side.
artefacts
stage 05
kingdom + phylum · plausibility · treatment
Each run emits kingdom and phylum composition, a plausibility metric (Bray-Curtis vs. the expected kingdom distribution), treatment-effect tests (Mann-Whitney U), and a top-features diagnostic table that flags RIE over-amplification. A MultiQC HTML report ties it together, and a provenance.yaml records git SHA, atlas DOI, container digests, and params.
artefacts
§ 03 · Run it
Try the demo
Download, run, inspect the exact input + output files the pipeline consumes. ~20 minutes on a laptop.
Inspect demo bundle# fetch demo bundle and run APPLY end-to-end curl -L -o demo_climgrass.tar.gz \ https://github.com/soillifeatlas/nf-core-soillifeatlas/releases/\ download/v0.1.0-dev/demo_climgrass.tar.gz tar -xzf demo_climgrass.tar.gz nextflow run soillifeatlas/nf-core-soillifeatlas \ -r v0.1.0-dev -profile docker \ --mode apply \ --atlas_path demo_climgrass/atlas \ --soil_intensity demo_climgrass/soil_intensity.parquet \ --soil_mgf demo_climgrass/soil.mgf \ --sample_metadata demo_climgrass/sample_metadata.tsv
Run on your data
Upload → metadata → instrument → advanced. You walk away with a copy-paste nextflow run command for your laptop, Docker, SLURM, or Google Batch.
Zero upload. The form is entirely client-side — your files never leave your browser. All it produces is a command line.
Hosted cloud
Eventually we'll run the pipeline on Google Batch for you and email the results. The google_batch profile ships as a stub today — execution unlocks when the grant lands.
§ 03.2 · Configure
The form below never contacts a server. It reads file sizes and names locally, composes a nextflow run command, and hands it back to you. Copy-paste into a terminal that has Nextflow installed and you're off.
v0.1 scope · mode=apply · POS ion mode · atlas=v0.1.0
§ 04 · Demo bundle
Twelve ClimGrass grassland samples, positive ion mode, already aligned. See exactly what the pipeline consumes and what it produces — every artefact is plain Parquet, TSV, HTML, or YAML.
inputs
4 items · 18 MB
consensus feature × sample matrix (MZmine 4.9.14 IIMN output)
MS2 consensus spectra, one per feature
treatment · replicate · ClimGrass plot id (optional)
v0.1.0 reference (simper_fp, RIE, IS masses, ArchLips)
outputs
6 items · ~1.6 MB
one Parquet per decomposition method
BC distance to expected kingdom composition
Mann-Whitney U per phylum × treatment
top 50 features driving each phylum (RIE guard)
MultiQC + custom panels
git_sha · atlas DOI · container digests · params
§ 05 · Hosted (soon)
Note · This form does nothing yet. The hosted service is scheduled for post-funding (see the Google.org Impact Challenge). In the meantime, the pipeline runs locally (Docker), on a SLURM cluster, or on any Nextflow-compatible compute.