Workflow Pipeline Execution Validation

workflow-pipeline-execution-validation

Summary

Validate input data files and confirm successful execution of a complete metabolomics annotation pipeline by checking intermediate and final outputs against expected structure and content. This skill ensures early detection of data issues and proper workflow completion before downstream analysis.

When to use

Apply this skill when you have prepared metabolomics input files (feature quantification table, MS/MS spectra in MGF format, sample metadata) and are about to execute the TIMA taxonomically informed annotation pipeline, or after pipeline execution to verify all outputs were generated correctly.

When NOT to use

• Input data is already in annotated metabolite format (not a raw feature table)
• MS/MS spectra file is missing or empty (.mgf contains no spectra records)
• Sample metadata does not link to organism taxonomy information (single organism analysis may proceed, but organism annotation will be limited)

Inputs

• Feature quantification table (.csv/.tsv with feature ID, retention time, m/z, sample intensity columns)
• MS/MS spectra file (.mgf format with fragment spectra)
• Sample metadata (.csv/.tsv linking samples to organisms)
• Optional: SIRIUS annotations (.zip v5/v6), GNPS-FBMN results, custom spectral libraries

Outputs

• Validated input report (spectra counts, feature counts, metadata consistency)
• Intermediate annotation outputs (SIRIUS results, spectral library matches, organism-specific assignments)
• Final annotation table with metabolite identities and confidence metrics
• Workflow execution log and intermediate data artifacts

How to apply

First, validate input files using the validate_inputs() function to check MGF spectra counts, feature table structure (presence of feature ID, retention time, m/z, and sample intensity columns), and metadata consistency before launching the pipeline. Then execute the canonical tima workflow pipeline via the Shiny GUI (tima::run_app()) or programmatically, specifying customizable column names and file paths. Finally, confirm workflow completion by verifying that all intermediate annotation outputs (SIRIUS results, GNPS-FBMN matches, spectral library matches) and final annotation tables are present with expected structure, non-empty content, and consistent feature identifiers across all outputs.

Related tools

• tima R package (Executes taxonomically informed metabolite annotation pipeline with integrated spectral, chemical, and taxonomic knowledge) — https://github.com/taxonomicallyinformedannotation/tima
• SIRIUS (Provides optional external metabolite structure predictions (v5/v6) that are integrated into annotation workflow)
• GNPS-FBMN (Provides optional spectral networking results that can be imported for annotation validation)
• LOTUS database (Default structure-organism pairs library (>650k pairs) used for organism-aware metabolite annotation) — https://lotusnprod.github.io/lotus-manuscript/
• Docker (adafede/tima-r image) (Containerized execution environment with all dependencies pre-installed for reproducible pipeline execution) — https://hub.docker.com/r/adafede/tima-r/

Examples

tima::validate_inputs(features="data/source/example_features.csv", spectra="data/source/example_spectra.mgf", metadata="data/source/example_metadata.tsv", feature_col="row ID", filename_col="filename", organism_col="ATTRIBUTE_species"); tima::run_app()

Evaluation signals

• validate_inputs() reports zero missing or malformed columns in feature table and spectra counts match non-zero value
• All intermediate output files present in expected directory structure with timestamps matching pipeline execution start time
• Final annotation table contains same number of features as input feature table with all feature IDs preserved
• No empty annotation records; each feature has at least one assigned structure candidate or metabolite identifier
• Metadata cross-reference integrity: all sample IDs in final output correspond to input metadata file and organism taxonomies are correctly assigned

Limitations

• Package lifecycle is experimental; changelog is not available per README, indicating potential API instability
• Workflow performance depends heavily on LOTUS database coverage for target organisms; organisms with sparse natural product data will yield lower annotation confidence
• Optional SIRIUS integration requires separate installation and licensing considerations (v5/v6 specified); workflow proceeds without it but with reduced structure prediction accuracy
• Column names must be customizable through Shiny app or YAML/CLI parameters; incorrect column name mapping will cause silent validation failures

Evidence

• [readme] Workflow validation methodology: "Start by validating your input files to catch issues early and save debugging time"
• [readme] Feature table schema requirements: "Must contain: feature ID, retention time, m/z, and sample intensity columns"
• [readme] Validation function implementation: "validate_inputs() will: Count spectra in MGF files, Count features and check required columns, Check metadata file consistency"
• [other] Pipeline execution workflow completion criteria: "Validate workflow completion by confirming all intermediate and final annotation outputs are generated with expected structure and content"
• [readme] LOTUS as default resource: "Structure-organism pairs library - We provide LOTUS (>650k pairs) as default"