Galaxy W4m Lcms Processing Workflow

galaxy-w4m-lcms-processing-workflow

The Workflow4Metabolomics LC-MS line as one pipeline: ten stages, each bound to leaf skills already in this collection and to the Galaxy tool that performs it.

When to use

You are running, reviewing or reproducing a metabolomics analysis on a Workflow4Metabolomics Galaxy instance, and you want the whole route from raw open-format files to statistics rather than one step of it.

When not to use

• The analysis is MS/MS annotation-led — molecular networking, spectral library matching, SIRIUS. That is untargeted-lcmsms-annotation, a different pipeline with different tools.
• You are working locally with xcms in R. The stage order still holds, but the Galaxy datatypes and the three-table interchange format do not apply.
• The data are NMR, or isotope-labelling and flux data. The same repository ships those tools, but they form separate chains that share only the three-table format.

Why this stage order

It is not a convention. Each Galaxy wrapper declares the datatype it accepts and the datatype it emits, and the chain below is the only order those declarations permit. Two consequences are worth naming because they surprise people: adjustRtime consumes a grouped object, so correspondence runs before alignment and is then repeated against the corrected retention times; and fillChromPeaks accepts only rdata.xcms.group, so gap filling cannot be moved after annotation.

Stages

import_raw

Raw open-format files -> an MSnbase object the xcms chain accepts.

• after: —
• in → out: mzML, mzXML, netCDF → rdata.msnbase.raw
• W4M tools: MSnbase readMSData, Mz(X)ML Shaper
• EDAM: operation_3215
• leaf skills:
  • xcms-data-import-preprocessing (primary)
  • mzml-file-import-xcms
  • mass-spectrometry-data-format-import
• grounding: 2 source DOI(s)

parameter_optimisation

Design-of-experiments search for peak-picking, grouping and alignment parameters.

• after: —
• in → out: mzML → ipo-parameters
• W4M tools: IPO for xcmsSet, IPO for group and retcor
• EDAM: operation_3435
• leaf skills:
  • parameter-tuning-metabolomics (primary)
  • peak-picking-parameter-configuration
  • chromatographic-alignment-parameter-selection
• grounding: 3 source DOI(s)

peak_detection

Detect chromatographic peaks per sample and merge the per-sample objects.

• after: import_raw, parameter_optimisation
• in → out: rdata.msnbase.raw → rdata.xcms.findchrompeaks
• W4M tools: xcms findChromPeaks, xcms refineChromPeaks, xcms findChromPeaks Merger
• EDAM: operation_3441
• leaf skills:
  • lc-ms-profile-data-segmentation (primary)
  • chromatographic-peak-detection-wavelet
  • chromatographic-profile-quality-assessment
• grounding: 3 source DOI(s)

correspondence

Group peaks across samples into features.

• after: peak_detection
• in → out: rdata.xcms.findchrompeaks → rdata.xcms.group
• W4M tools: xcms groupChromPeaks
• EDAM: operation_3632
• leaf skills:
  • lc-ms-feature-grouping-by-retention-time (primary)
  • mass-chromatogram-alignment
• grounding: 1 source DOI(s)

rt_alignment

Correct retention-time drift, then repeat correspondence against the corrected times.

• after: correspondence
• in → out: rdata.xcms.group → rdata.xcms.group
• W4M tools: xcms adjustRtime, xcms groupChromPeaks
• EDAM: operation_3632
• leaf skills:
  • retention-time-correction-and-alignment (primary)
  • retention-time-alignment-correction
• grounding: 2 source DOI(s)

gap_filling

Integrate the areas of peaks missing from the feature matrix.

• after: rt_alignment
• in → out: rdata.xcms.group → rdata.xcms.fillpeaks
• W4M tools: xcms fillChromPeaks
• EDAM: operation_3214
• leaf skills:
  • missing-peak-imputation-fillpeaks (primary)
• grounding: 1 source DOI(s)

annotation

Group features into pseudospectra and annotate isotopes and adducts.

• after: gap_filling
• in → out: rdata.xcms.fillpeaks → rdata.camera, three-table
• W4M tools: CAMERA.annotate, CAMERA.groupFWHM, CAMERA.groupCorr, CAMERA.findIsotopes, CAMERA.findAdducts, CAMERA.combinexsAnnos
• EDAM: operation_3432
• leaf skills:
  • spectral-feature-clustering-by-intensity-correlation (primary)
  • metabolite-mass-to-charge-ratio-matching
  • adduct-fragment-table-construction
• grounding: 1 source DOI(s)

table_conformance

Make the dataMatrix, sampleMetadata and variableMetadata agree before any statistics.

• after: annotation
• in → out: three-table → three-table
• W4M tools: Check Format, Table Merge, W4M concatenate, xcms get a sampleMetadata file
• EDAM: operation_3891
• leaf skills:
  • feature-metadata-parsing-and-integration (primary)
  • metadata-structure-checking
  • cross-table-metadata-harmonization
• grounding: 3 source DOI(s)

normalisation

Normalise intensities and correct batch and drift effects using the pooled QC samples.

• after: table_conformance
• in → out: three-table → three-table
• W4M tools: Normalization, Parsec batch correction, Batch Dispersion, Intensity Check
• EDAM: operation_3434
• leaf skills:
  • batch-effect-correction-qc-reference (primary)
  • batch-effect-correction-and-adjustment
  • compound-reliability-rsd-filtering
• grounding: 3 source DOI(s)

statistics

Univariate, multivariate and signature analysis on the corrected matrix.

• after: normalisation
• in → out: three-table → statistics, pdf
• W4M tools: Biosigner, mixmodel, Metabolites Correlation Analysis, Heatmap
• EDAM: operation_3695
• leaf skills:
  • pls-pls-da-classification-modeling (primary)
  • metabolomic-biomarker-pathway-association
• grounding: 2 source DOI(s)

Grounding

Every stage carries the DOIs and KB slugs of the leaves bound to it; ground a stage with the collection's /ground recipe against those slugs.

The pipeline structure is grounded differently from the leaves: it was read off the tool wrappers in workflow4metabolomics/tools-metabolomics at master on 2026-08-21, not distilled from a paper. provenance.structure_source in workflow.yaml records that.

Verification contract

• Each stage's output datatype is the one the next stage declares as its input. A Galaxy invocation that type-checks has already verified most of this.
• The second correspondence pass ran after alignment; a workflow with one grouping step is missing it.
• The three tables agree on identifiers and order before normalisation. See w4m-three-table-format-conformance.
• xcms process history matches the exported workflow.
• Feature counts are reported after gap filling, not after correspondence; the two differ and only the first describes the matrix the statistics saw.

Coverage gaps

The corpus does not cover this pipeline evenly, and the bindings say where:

• Parameter optimisation — no leaf declares IPO. The stage is bound to generic parameter-tuning leaves, which describe the activity but not the tool.
• Statistics — no leaf declares Biosigner or the other W4M statistics wrappers. The bound leaves cover PLS-DA and biomarker association generically.
• The platform — no leaf declares Workflow4Metabolomics or Galaxy. The collection knows the methods this pipeline runs and not the platform that runs them.

Of the 16 DOIs the wrappers cite, 2 are already in the collection's corpus. The other 14 are candidates for corpus expansion, which would close the gaps above at their source rather than by rebinding.

Provenance

Structure from the tool wrappers; leaf bindings by lexical retrieval over skills_index.json constrained to leaves that declare the relevant tool, then checked by hand. bound_by: index. No ASB rerun; no per-paper workflow DAG contributed structure, so derived_from_workflows is empty and no benchmark ablation is required.