PaperlayerView exact revision on arXiv

Source-linked AI summary

Create, run, share, publish, and reference your LC-MS, FIA-MS, GC-MS, and NMR data analysis workflows with the Workflow4Metabolomics 3.0 Galaxy online infrastructure for metabolomics

Yann Guitton, Marie Tremblay-Franco, Gildas Le Corguillé, Jean-François Martin, Mélanie Pétéra, Pierrick Roger-Mele, Alexis Delabrière, Sophie Goulitquer, Misharl Monsoor, Christophe Duperier, Cécile Canlet, Remi Servien, Patrick Tardivel, Christophe Caron, Franck Giacomoni, Etienne Thévenot

arXiv:1708.04024v1q-bio.QM

TL;DR

Complex metabolomics data, diverse experimental designs, and numerous analysis tools make comprehensive, rigorous analysis challenging. W4M 3.0 combines tools for several metabolomics technologies and demonstrates them in three case studies, with analyses that can be referenced through permanent DOIs.

  • Problem

    Metabolomics produces complex, high-volume data and involves diverse experimental designs and tools, creating a need for user-friendly, modular, computationally efficient analysis platforms.

    Source: §1 · p. 5, §4 · p. 28

  • Method

    W4M 3.0 provides an online Galaxy-based infrastructure with 40 tools for LC-MS, FIA-MS, GC-MS, and NMR analysis and applies it to three published case studies.

    Source: §1 · p. 8, §4 · p. 28

  • Results

    The three case studies demonstrate comprehensive analyses across LC-MS, GC-MS, and NMR; algal PLS-DA discrimination had Q2Y = 0.84.

    Source: §1 · p. 8, §3.1.2 · p. 22

  • Takeaways & Limitations

    Complete analyses, including workflows, parameter values, input data, and outputs, can be publicly referenced with a permanent DOI for reproduction.

    Source: §4 · p. 30

Abstract

from arXiv · show

Metabolomics is a key approach in modern functional genomics and systems biology. Due to the complexity of metabolomics data, the variety of experimental designs, and the variety of existing bioinformatics tools, providing experimenters with a simple and efficient resource to conduct comprehensive and rigorous analysis of their data is of utmost importance. In 2014, we launched the Workflow4Metabolomics (W4M, http://workflow4metabolomics.org) online infrastructure for metabolomics built on the Galaxy environment, which offers user-friendly features to build and run data analysis workflows including preprocessing, statistical analysis, and annotation steps. Here we present the new W4M 3.0 release, which contains twice as many tools as the first version, and provides two features which are, to our knowledge, unique among online resources. First, data from the four major metabolomics technologies (i.e., LC-MS, FIA-MS, GC-MS, and NMR) can be analyzed on a single platform. By using three studies in human physiology, alga evolution, and animal toxicology, we demonstrate how the 40 available tools can be easily combined to address biological issues. Second, the full analysis (including the workflow, the parameter values, the input data and output results) can be referenced with a permanent digital object identifier (DOI). Publication of data analyses is of major importance for robust and reproducible science. Furthermore, the publicly shared workflows are of high-value for e-learning and training. The Workflow4Metabolomics 3.0 e-infrastructure thus not only offers a unique online environment for analysis of data from the main metabolomics technologies, but it is also the first reference repository for metabolomics workflows.

Graphical abstract

The listed keywords identify metabolomics, data analysis, e-infrastructure, workflow, Galaxy, and repository. The abbreviations cover metabolomics technologies and related analysis terms.

Highlights

  • The keywords are metabolomics, data analysis, e-infrastructure, workflow, Galaxy, and repository.Source: Highlights · p. 3

Abbreviations:

  • DOI means Digital Object Identifier, and e-infrastructure means online infrastructure.Source: Abbreviations · p. 3
  • LC means Liquid Chromatography, MS Mass Spectrometry, NMR Nuclear Magnetic Resonance, and FIA Flow Injection Analysis.Source: Abbreviations · p. 3
  • GC means Gas Chromatography, HR High-Resolution, and QC Quality control sample (pool of all samples).Source: Abbreviations · pp. 3–4
  • PCA means Principal Component Analysis, (O)PLS-DA (Orthogonal) Partial Least Squares - Discriminant Analysis, and SVM Support Vector Machine.Source: Abbreviations · p. 4
  • VIP means Variable Importance in Projection, and W4M means Workflow4Metabolomics.Source: Abbreviations · p. 4

1. Introduction

Metabolomics analysis combines many tasks, tools, and parameters, making analyses time-consuming and difficult to reproduce when software and workflows are handled separately. W4M 3.0 extends Galaxy-based workflow analysis with tools for multiple metabolomics technologies and online referenceable analyses.

1. Introduction

  • Metabolomics analysis spans preprocessing, statistics, and annotation, with multiple alternative tasks, tools, and parameters making analysis combinations numerous.Preprocessing generates a sample-by-variable intensity matrix; statistical analysis detects variables of interest and builds prediction models, while annotation investigates chemical and biological functions.Source: §1 · pp. 5–6
  • Sequential use of varied software is time-consuming, repetitive, and error-prone, while unsaved workflows and parameter values impede efficient, reproducible analysis.Source: §1 · pp. 5–6
  • Graphical workflow systems let users select tools and parameters, chain tasks, and focus on analysis design without managing installation, scripts, or data formats.Source: §1 · p. 7
  • W4M initially integrated Galaxy with 20 tools for LC-HRMS workflows, alongside high-performance computing, online documentation, and a help desk.Source: §1 · p. 8
  • W4M 3.0 adds 20 tools for LC-MS, GC-MS, FIA-MS, and NMR workflows, and its complete histories can be referenced online with a permanent DOI.The release is presented with three case studies whose workflow outputs are compared with published results.Source: §1 · p. 8

2. Case Studies

Three published studies illustrate different biological questions, experimental designs, and metabolomics technologies. They cover human urine LC-HRMS, brown-alga GC-MS, and mouse-brain NMR analyses.

2. Case Studies

  • The case studies were selected to illustrate diversity in biological questions, experimental designs, and analytical technologies.Source: §2 · p. 10

2.1. Sacurine human physiological study (LC-HRMS)

  • The Sacurine studies used LC-HRMS on urine from 184 volunteers to examine metabolite identification and concentration variation with age, BMI, and gender.Raw data were preprocessed with XCMS and annotated with CAMERA; selected metabolites were matched against KEGG, HMDB, and METLIN, with MS/MS interpretation and refined intensity quantification.Source: §2.1 · p. 10

2.2. Ectocarpus (brown algae) ecological study (GC-MS)

  • The Ectocarpus study compared a freshwater strain cultured at medium or seawater salinity with a seawater strain cultured in seawater.Algal profiles were analyzed by GC-MS after harvesting and preparation that included drying, grinding, methanol extraction, and addition of ribitol as an internal standard.Source: §2.2 · p. 11

2.3. Bisphenol A Mus musculus toxicological study (NMR)

  • The mouse toxicology study assessed perinatal BPA exposure in male offspring brain samples using 1H NMR spectra acquired with a CPMG spin-echo sequence.Mothers received either 0.025 or 0.25 µg BPA/kg body weight/day during gestation and lactation; spectra were phased and baseline corrected.Source: §2.3 · pp. 11–12

3. Results

The results describe W4M 3.0’s metabolomics tools across LC-HRMS, FIA-MS, GC-MS, and NMR, with case studies demonstrating analysis workflows and NMR annotation of metabolites associated with BPA exposure. They also present W4M as a DOI-based repository for sharing and reusing metabolomics analysis histories.

3. Results

  • The W4M 3.0 release provides 40 tools, including 20 new modules for advanced LC-HRMS, FIA-MS, GC-MS, and NMR workflows; the 18-tool Sacurine-statistics workflow also uses generic statistical tools applicable across omics data.Source: §3 · p. 12, §3.1 · p. 13

3.1. Analysis of LC-MS, GC-MS, and NMR metabolomics data sets on W4M

  • W4M statistical results matched the publication, producing a peak table of 183 samples and 109 metabolites after quality control removed one sample and four variables.The history was publicly shared as W4M00001 with a permanent DOI.Source: §3.1 · pp. 14–15
  • centWave detected 372 previously characterized urinary metabolites versus 342 with the original settings, yielding 3,120 QC-filtered features including 85% of matchedFilter ions, and was selected for the reference workflow.Source: §3.1 · p. 17
  • Biosigner selected eight features; PLS-DA, Random Forest, and SVM achieved classification performances of 84%, 90%, and 92%, respectively, including three previously identified metabolites and four additional high-performing ions.Source: §3.1 · pp. 17–18
  • The new metaMS.runGC tool groups correlated GC-MS ions into compound-level pseudospectra, and Golm/NIST matching correctly identified citric acid, confirmed by injection of the pure compound.The new Golm search tool also supports exported spectra for querying NIST and reuse as an in-house database.Source: §3.1.2 · pp. 18, 19, 21
  • Reference-spectrum reprocessing recovered mannitol in every sample after retention-time shifts had caused misses; PCA separated the three algal cultures, and PLS-DA achieved significant Q2Y = 0.84.The lower mannitol concentrations in the freshwater strain agreed with Dittami et al. (2012).Source: §3.1.2 · pp. 21–22
  • The NMR OPLS-DA model significantly discriminated the two BPA doses and selected 157 buckets with VIP > 0.8, while no univariate differences remained significant after 5% FDR correction.PCA identified two observations for exclusion before subsequent analyses.Source: §3.1.3 · p. 24
  • NMR annotation identified metabolites associated with BPA exposure and found Taurine and Aspartate to be additional discriminative neurotransmitters.The workflow annotated 39 metabolites. Glutamic Acid had VIP 4.1 for the 2.35 ppm bucket and GABA had VIP 2.6 for the 2.3 ppm bucket, confirming reported decreases after BPA exposure (Cabaton et al., 2013); Taurine had VIP 2.9 for the 3.26 ppm bucket and Aspartate had VIP 2.2 for the 3.90 ppm bucket.Source: §3.1.3 · p. 25

3.2. Referencing histories

  • W4M published case-study histories with DOIs, allowing users to import histories and extract workflows for application to new data sets.The histories include workflows, input data, and output results; W4M is described as the first repository dedicated to metabolomics data-analysis workflows.Source: §3.2 · pp. 25–26

4. Discussion

W4M 3.0 combines comprehensive metabolomics workflows across major technologies with public referencing and sharing. Case studies show that these analyses can be run efficiently and produce findings consistent with prior results while suggesting additional observations.

4. Discussion

  • The three case studies confirmed published findings, selected and annotated key metabolites, and suggested neurotransmitter changes even at the lowest BPA exposures.The Sacurine-comprehensive workflow used 29 tools on 234 raw files in a few hours.Source: §4 · p. 28
  • W4M 3.0 offers 40 tools across LC-MS, FIA-MS, GC-MS, and NMR, including original tools and parallel processing for computationally intensive LC-MS preprocessing.The release doubled the available tools compared with the previous version.Source: §4 · pp. 28–29
  • W4M provides a single online resource for comprehensive analysis across LC-MS, FIA-MS, GC-MS, and NMR, with Galaxy workflow-management features for building, running, and sharing workflows.The authors contrast this scope with online platforms focused on LC-MS processing and annotation or statistical analysis.Source: §4 · pp. 29–30
  • Public histories receive a permanent DOI and expose workflows, data, parameters, and results for citation and reproducibility, positioning W4M as a metabolomics workflow repository.The authors describe this as filling a gap between raw-data repositories and spectral or metabolite databases.Source: §4 · p. 30
  • Tutorials, reference histories, and internet-accessible hands-on sessions support e-learning and training in workflow design and interpretation.The authors report regular one-week courses combining practical presentations and tutoring.Source: §4 · pp. 30–31
  • The authors conclude that W4M 3.0 provides features to learn, design, run, share, and reference comprehensive metabolomics analyses.Source: §4 · p. 31

Supplementary material

The supplementary material includes an author list and a Table S3 entry for the BPAMmusculus workflow.

Supplementary material

1. Workflow Management with W4M

W4M workflow management covers data preparation and upload, workflow construction and execution, and sharing or publishing histories. Publicly referenced histories make the analysis inputs, settings, and outputs accessible to reviewers and readers.

1. Workflow Management with W4M

  • MS raw files should use mzML, mzXML, mzData, or NetCDF; centroid data are recommended to reduce file size, with conversion performed before upload.Source: §1.2.1.1 · p. 40
  • Raw files should be organized by class because class information is used by xcms.group with minfrac to filter features not detected in enough samples in at least one class.Raw files should be zipped before upload.Source: §1.2.1.1 · p. 40
  • For LC-MS, individual files can be grouped as a data collection for parallel xcms.xcmsSet processing; resulting xset.RData files and sample metadata are merged before xcms.group.The text states that using a data collection speeds this computationally intensive step.Source: §1.2.1.1 · pp. 40–41
  • Preprocessed data use three tabulated files—dataMatrix, sampleMetadata, and variableMetadata—a format used by downstream normalization, quality-control, statistical-analysis, and annotation tools.Source: §1.2.2 · p. 41
  • Files up to 2 GB can be uploaded by drag and drop, while larger datasets require an FTP client; NMR sample directories require Bruker files with acquisition and processing numbered 1, gathered under a zipped parent directory.Source: §1.3 · pp. 41–42, §1.2.1.2 · p. 41
  • Users can build workflows by chaining tools and selecting parameters in the canvas, or extract a workflow from an existing history.Source: §1.4 · p. 42
  • W4M provides browser access to 4,000 computing cores without using local resources; histories can be shared with named users or published openly, and tutorials cover data preparation, upload, analysis, sharing, and publishing.Source: §1.5 · p. 42, §1.6 · p. 42, §1.8 · p. 42
  • Published histories can be cited by reference ID and permanent DOI, giving reviewers and readers access to the workflow, parameters, input and output data, and metadata.Source: §1.7 · p. 42

2. Case Studies workflows

The case-study workflows provide DOI-linked histories that can be imported and extracted in Galaxy, with tools and critical points described for three studies. Their procedures span sample quality control, preprocessing, normalization, statistical analysis, and metabolite annotation.

2. Case Studies workflows

  • The reference workflows are available through permanent DOI links; users can import each history into Galaxy and extract its workflow for reuse.The paper describes the tools, parameter values, and critical points for each case-study workflow.Source: §2 · p. 43

W4M00002_Sacurine-comprehensive

  • Sacurine metadata supplies sample type, injection order, batch, osmolality, age, BMI, and gender for quality control, normalization, and statistical analyses.The described univariate tests assess Spearman correlation with age or BMI and median differences between genders.Source: §2 · pp. 44–45
  • Sacurine processing combines quality filtering, signal-drift and batch correction, and osmolality normalization; loess correction requires at least five pool injections per batch.The correction model should be checked graphically, and pool dilution correlations can serve as an additional quality metric when dilutions are available.Source: §2 · p. 45

W4M00004_GCMS-Algae

  • The GC-MS algae workflow processes 12 converted Agilent files with metaMS.runGC and searches spectra against the Golm Metabolome Database.The workflow notes that file splitting affects xcms.group minfrac; overly high similarity thresholds may misalign low-intensity compounds, so pseudospectra should be checked graphically.Source: §2 · p. 47
  • The algae workflow also supports local NIST searching and reprocessing with a user-created spectral database containing named citric acid, mannitol, and ribitol spectra.The modified .msp database is uploaded into Galaxy for reprocessing.Source: §2 · p. 47

W4M00006_BPAMmusculus

  • The BPA mouse NMR workflow buckets and integrates spectra, normalizes to total intensity, and models treatment response with pareto-scaled OPLS-DA.Exclusion zones depend on the biological matrix and sample-preparation solvents; permutation diagnostics use pQ2 < 0.05 as a check against overfitting.Source: §2 · p. 49
  • The BPA mouse workflow assesses treatment-dose differences with Wilcoxon tests and false-discovery-rate correction, after filtering variables with Treatment_OPLSD A_VIP_pred < 0.8.Its LC-MS feature-processing steps include retention-time alignment, peak grouping, missing-value imputation, and CAMERA annotation of possible isotopes, adducts, and fragments.Source: §2 · pp. 44, 49
Loading 1708.04024v1…