eduKateSG · Why Science?
A metabolomics result becomes interpretable when study, sample, assay, protocol and metabolite evidence stay connected
Follow a MetaboLights study from MTBLS accession through ISA metadata, factors, assays, raw spectra, derived data and metabolite assignments before comparing abundance or identity claims.
Metabolomics studies small molecules and chemical features in biological systems. The final table can look deceptively simple: names or features in rows, samples in columns and numbers in cells. MetaboLights helps preserve the study description, samples, assays, protocols, raw data, derived files and metabolite information needed to interpret those numbers.
The resource uses linked ISA-style metadata so a reader can move from an MTBLS study accession to biological material, analytical workflow and reported metabolite assignments. This guide turns that structure into a cheerful, careful science-learning journey.
Build four objects as you read: a study card, sample-factor table, sample-to-file map and metabolite evidence table. The aim is to let another student trace one reported value all the way back to the biological source and measurement.
Choose a small public study for the first exercise and write the scientific question before downloading files. A good question names the biological system, conditions and measurement. It also states whether the aim is to reproduce a submitted comparison, explore processed values or reanalyse raw spectra.
Keep archive files unchanged. Store the original metadata and data in a source directory, make a manifest with filenames and hashes where possible, and create a separate working directory for cleaned tables. This simple layout prevents a convenient edit from replacing the deposited evidence.
Metabolomics is sensitive to preparation and analytical context. Extraction solvent, storage, chromatography, ionisation, polarity, NMR settings and batch can change which features are observed. Record the fields relevant to the chosen comparison rather than assuming every abundance value has the same measurement meaning.
Quality-control samples help reveal instrument drift and processing stability. Blanks help identify background or contaminants. Standards can support retention or chemical identity. These rows should remain in the design map even when the biological statistical model excludes them.
Feature alignment and peak picking create derived objects. A single compound can generate several ions or adducts, and one feature can remain ambiguous. Count features and named metabolites separately so the report does not turn computational rows into a list of unique chemicals.
Missing values deserve method-aware treatment. Absence may reflect a concentration below detection, failed peak assignment, filtering or a data-processing choice. Record the study's convention and any imputation rule; never translate every blank cell into biological absence.
Batch correction and normalisation can improve comparison, but they also transform the values. Save diagnostics, formulas and software versions, and check whether treatment groups are confounded with analytical batches. No correction method can recover a comparison that the design never separated.
Chemical identifiers are especially helpful when names vary. Preserve database accessions, formula, exact mass and structural identifiers when provided. Common names can refer to families or stereoisomers, so identity claims should stay at the level supported by the study.
Plan the graph after the evidence table. Show individual biological samples where possible, label units or transformed scale and distinguish quality-control objects. A heatmap can reveal patterns, but it should not hide sample count, unknown features or confidence boundaries.
The final conclusion should have three layers: what the submitter reported, what the downloaded values show under your stated method and what biological explanation remains plausible. Separating those layers keeps reuse honest and makes follow-up experiments easier to design.
Before importing tables, inspect delimiters, headers, decimal conventions and character encoding. Spreadsheet software can alter long identifiers, formulas or dates. Keep a text-preserving copy and verify row and column counts after every conversion.
Create a data dictionary for every analysis column. Include field name, source file, unit, allowed values and any recoding rule. This gives factor labels, chemical identifiers and numerical measurements a common reference and prevents two similarly named columns from being treated as equivalent.
Run an audit join from sample sheet to assay table. Require each selected sample to match the expected assay rows and list extras separately. Multiple assays can be legitimate, but the relationship must be represented explicitly before values are combined.
Build a file manifest with raw, derived, metadata and supplementary categories. Compare the filenames referenced in assay tables with the downloadable objects and record missing or unreferenced files. A study page can be informative even when a local analysis should pause for incomplete linkage.
For mass-spectrometry studies, preserve ion mode, adduct assumptions and feature identifiers. Positive- and negative-ion measurements may detect related chemistry through different signals. Combining them without a documented rule can double-count or misidentify compounds.
For NMR studies, record nucleus, pulse-sequence and spectral-processing context supplied by the study. Peak positions and integrals depend on acquisition and processing choices. The same study-level repository can therefore hold evidence that needs method-specific interpretation.
Retention time can support feature alignment or identity, but values are method dependent. Do not compare retention values across chromatography systems without examining conditions and standards. Keep the unit and protocol reference beside the number.
Mass accuracy and database matching narrow possibilities but may not distinguish isomers. If several compounds share a formula or mass, keep the candidate set visible. Strong chemical names should appear only when the evidence level justifies them.
When a paper reports pathways or biological mechanisms, trace those conclusions back through metabolite assignments and statistics. Enrichment results inherit uncertainty from feature detection, identification and background choices. Cite the pathway method separately from the repository files.
Statistical analysis needs the biological sample count, not merely the number of injections or features. Show group sizes, effect estimates and uncertainty, and correct for multiple testing when many features are assessed. A small p-value cannot repair weak identity or confounded design.
Visualisation should preserve provenance. Label axes with units or transformation, identify whether values are submitter processed or recomputed, and make sample exclusions visible. Colour palettes can aid interpretation but should not substitute for a legend or text description.
Use a stopping rule when critical identity or sample metadata are absent. The scientifically useful output may be a data-quality report describing what can and cannot be analysed. This prevents an attractive plot from giving uncertain inputs false authority.
For peer review, give another student the MTBLS accession, one sample, one assay row, one data file and one metabolite row. Ask them to reconstruct the route and the confidence boundary. Any broken link identifies exactly where the project needs better documentation.
Finally, store the source archive date, software environment and derived tables together. A future learner can rerun the method against the same study revision or deliberately compare a later revision without confusing the two.
Comparing several MetaboLights studies requires a second harmonisation layer. Align organism, tissue, factor, analytical method and chemical identifiers before stacking abundance tables. Preserve study-specific units and processing, and avoid batch correction across studies unless the design and method justify it.
Negative evidence needs careful language. A metabolite not reported in a table may fall below detection, remain an unnamed feature, be excluded during processing or lie outside the analytical method. Say “not reported under this study workflow” before claiming biological absence.
Data licences and ethical conditions also matter. Public access supports reuse, while human-derived studies still require responsible interpretation and respect for stated policies. Classroom projects should focus on group-level scientific questions and avoid personal or clinical conclusions unsupported by the dataset.
A compact claim ledger can join statement, samples, assay, files, metabolite row, analysis and limitation. Review that ledger before writing the abstract. If a claim lacks one of these links, either repair the evidence trail or narrow the sentence.
These habits make complex chemistry approachable. Students do not need to memorise every instrument setting; they need to know which settings affect the question, where they are recorded and how uncertainty flows into the final interpretation.
That is the practical heart of careful, reusable metabolomics evidence for future scientific learning and honest evidence comparison.
Chapter index
Chapters 1–9: from MTBLS identity to experimental design
- 1. Cite the permanent MTBLS accession
- 2. Check study status and release context
- 3. Read the investigation and protocols first
- 4. Build the sample-factor table
- 5. Separate experimental samples and controls
- 6. Count replication at the correct level
- 7. Follow samples into assay files
- 8. Distinguish analytical protocols
- 9. Separate raw, derived and metabolite files
Chapters 10–17: from assay files and metabolite tables to bounded reuse
Public MetaboLights studies use MTBLS accessions. Temporary REQ identifiers belong to the submission workflow and should not be used as the permanent citation for a public study.
Try this
Copy the MTBLS identifier, title and public URL into the project notebook. Test that the accession retrieves the intended study and compare it with the manuscript's data-availability statement.
Keep the accession in derived filenames and figure captions. A local nickname such as “metabolomics study” cannot reconnect a result to the live archive.
Evidence checkpoint
Record the permanent MTBLS accession rather than a temporary REQ identifier when citing a public study.
CHAPTER 2 OF 17 · Identify study and release
2. Check study status and release context
Read the study's public status, release information, submitter or contact fields and linked publication. These fields establish scope and timing, especially when metadata or files have been revised.
Try this
Compare title, authors, organism and experimental aim with the publication. Record any difference in sample count or assay description rather than assuming every document was frozen at the same moment.
An archive update can improve metadata without invalidating earlier work. Date your retrieval so later readers can compare versions fairly.
Evidence checkpoint
Check study status, release context, title and linked publication before reusing files.
CHAPTER 3 OF 17 · Identify study and release
3. Read the investigation and protocols first
The investigation description provides title, contacts, abstract, publications, design descriptors and protocol references. It frames the biological question before any abundance table is opened.
Try this
Rewrite the study design in one sentence naming organism or system, intervention or condition, analytical method and intended comparison. Mark missing information instead of filling it from expectation.
Then list the protocols in order. This reveals where sampling, extraction, chromatography, mass spectrometry, NMR or data transformation enter the evidence chain.
Evidence checkpoint
Read the investigation description and protocols before opening abundance tables.
MetaboLights sample sheets assign unique sample names and describe organism, organism part, sample type and study factors. These rows define the biological objects being compared.
Try this
Create one row per sample and preserve original field values. Add clean analysis labels in new columns rather than overwriting source terms. Count each factor combination and compare it with the stated design.
If a factor is absent or inconsistent, record the limitation. A filename or numerical pattern cannot safely invent biological metadata.
Evidence checkpoint
Build a sample table from unique sample names, organisms, organism parts, sample types and study factors.
CHAPTER 5 OF 17 · Read samples and factors
5. Separate experimental samples and controls
Metabolomics studies may include experimental samples, biological controls, solvent blanks, extraction blanks, pooled quality-control samples and standards. Each serves a different purpose.
Try this
Classify sample type from metadata and protocols. Plotting blanks as if they were biological replicates or merging pooled quality controls into a treatment group changes the experiment.
Keep controls visible in the design table even when they are excluded from a biological comparison. Their presence can explain contamination checks, drift correction or feature filtering.
Evidence checkpoint
Keep controls, blanks, quality-control samples and experimental samples distinct.
One biological source can produce several preparations, assays or instrument injections. These repeated files improve technical assessment but do not automatically increase biological sample size.
Try this
Trace sample names into assay rows and add identifiers for extraction, batch and injection where available. Count biological units, preparation replicates and repeated measurements separately.
When pooling occurs, state which sources contributed. A pooled quality-control file represents mixture performance, not an independent organism or patient.
Evidence checkpoint
Separate biological replication from repeated injections, pooled samples and technical preparation steps.
Assay metadata connects sample names to protocol steps, raw spectral files, derived files and metabolite assignment tables. This join is the bridge between biology and instrument data.
Try this
Select one sample and follow the row across every file reference. Then start from a raw filename and trace backward to the sample. Both directions should agree.
Count unmatched names and duplicated links. Investigate them before analysis because a mismatch can attach a chemical profile to the wrong condition.
Evidence checkpoint
Follow each sample name into the assay file and verify its raw and derived data filenames.
Chromatography, mass spectrometry and NMR answer different measurement questions and use different metadata. Even within mass spectrometry, ionisation, polarity and acquisition choices matter.
Try this
Build a protocol timeline for each assay. Record separation method, instrument, acquisition mode and relevant settings without treating one brand or platform as a universal quality score.
Compare only samples processed under compatible conditions unless the analysis explicitly models batch or platform differences. Method context belongs beside every quantitative comparison.
Evidence checkpoint
Distinguish chromatography, mass spectrometry and NMR protocol fields where relevant to the assay.
CHAPTER 9 OF 17 · Trace assays and files
9. Separate raw, derived and metabolite files
Raw spectral files preserve instrument output. Derived files may contain peaks, aligned features or processed abundance. Metabolite assignment files connect features to reported chemical identities or unknowns.
Try this
Label each file by role in a manifest. Do not call a peak table raw because it is close to the instrument, and do not call a metabolite table confirmed merely because it contains names.
Keep filename, format, size and checksum where available. Reuse begins with knowing exactly which objects were downloaded and which stage they represent.
Evidence checkpoint
Treat raw spectra, derived peak tables and metabolite assignment files as different evidence objects.
Metabolomics values can be peak areas, intensities, ratios, concentrations or transformed and normalised measures. Magnitude alone cannot reveal the unit or processing history.
Try this
Read protocol and derived-data documentation before comparing cells. Record scaling, normalisation, log transformation and missing-value treatment supplied by the submitter.
If you transform values, preserve the original table and write the formula. A later reader should be able to reproduce the derived number or return to the deposited value.
Evidence checkpoint
Record units, transformations, normalisation and missing-value handling before comparing reported quantities.
Not every detected feature has a confident metabolite identity. Some rows represent mass-to-charge and retention features, unresolved compounds or ambiguous assignments. Renaming them all as known metabolites erases uncertainty.
Try this
Create separate columns for feature identifier, reported metabolite name, unknown status and confidence information. Keep duplicate or isomeric possibilities rather than selecting the most familiar compound.
Unknown features can still show reproducible patterns. The scientific claim should match the evidence: a feature changes, even if its chemical identity remains unresolved.
Evidence checkpoint
Separate measured features from metabolite names and keep unknown or ambiguous features visible.
CHAPTER 12 OF 17 · Trace assays and files
12. Evaluate metabolite identification evidence
A metabolite name may be supported by exact mass, retention information, fragmentation, standards, database matching or a combination. These evidence levels are not interchangeable.
Try this
Inspect database references, chemical identifiers, mass-to-charge values and confidence fields in the assignment table. Read the linked publication for additional validation procedures.
Use verbs such as annotated, putatively identified or confirmed according to the supplied evidence. Avoid upgrading a database match into confirmation without a suitable standard or method.
Evidence checkpoint
Inspect identifiers, database references, mass-to-charge values, retention information and confidence fields before asserting chemical identity.
CHAPTER 13 OF 17 · Evaluate metabolite claims
13. Preserve controlled terms and free text
MetaboLights encourages controlled vocabularies and ontology terms for searchable metadata. A term can improve consistency while the submitter's free text carries extra detail.
Try this
Save both label and ontology identifier when available. Do not replace an unresolved term with a near match merely to make the column uniform.
The Ontology Lookup Service guide shows how to inspect definitions and mappings before harmonising studies.
Evidence checkpoint
Use controlled vocabulary or ontology terms as defined metadata, while preserving submitter free text and unresolved mappings.
Submission validation checks required metadata, file references and structural rules. Passing validation improves consistency and reuse, but it does not independently verify every metabolite identity or biological interpretation.
Try this
Read errors and warnings where reports are available. Treat them as archive workflow evidence and inspect the scientific methods separately.
This distinction encourages fair evaluation. A well-structured dataset can support reanalysis even when a hypothesis remains uncertain, and a compelling paper still needs reusable metadata.
Evidence checkpoint
Treat validation reports as checks on submission structure and required metadata, not proof of every biological conclusion.
Use the MTBLS accession for the study, the paper for scientific interpretation and filenames or checksums for exact analysis objects. Place each citation beside the claim it supports.
Try this
When only one assay or subset is used, list the included samples and files. A broad study citation cannot identify which measurements entered a graph.
Add access and analysis dates. Public studies can receive metadata updates or new links while retaining their accession.
Evidence checkpoint
Cite the MTBLS study, linked paper and the exact files or tables used in your analysis.
Choose one metabolite or feature row and trace it through the assignment file, assay row, raw or derived file and sample-factor table. Record each join key.
Try this
Ask a partner to follow the same path using only your identifiers and notes. Differences in sample names, assay selection or identity confidence reveal where documentation needs repair.
The exercise turns a colourful abundance plot into an inspectable evidence chain. It also teaches why data organisation is part of chemistry, not paperwork after chemistry.
Evidence checkpoint
Ask a partner to trace one result from metabolite row through assay file and sample sheet back to the study factor.
CHAPTER 17 OF 17 · Evaluate metabolite claims
17. Connect metabolomics to future learning
MetaboLights combines chemistry, biology, instrumentation, statistics and data stewardship. It can inspire subject and career exploration, but one archive exercise cannot promise grades, admission or employment.
Try this
Continue through the Science Learning Hub and use the Education Hub to choose a realistic next capability such as chemical structure, experimental design, graphing or coding.
Did You Know? A metabolite table is the end of many linked decisions. Keeping study, sample, assay and identity evidence connected makes every number more meaningful and every limitation easier to explain.
Evidence checkpoint
Connect chemistry, biology, statistics and data provenance to a next learning step without promising an outcome.
Contents · Previous chapter · Continue at the Science Learning Hub
A final MetaboLights reading routine
Before analysing a public study, save the MTBLS accession, release or revision context, investigation file, sample sheet, assay files, metabolite assignment files and data manifest. Count the sample names and confirm that every selected data file joins to the intended sample and assay.
Finish by separating detected feature, annotated metabolite and confirmed chemical identity. Record the identification fields and confidence supplied by the study, then state which interpretation comes from the submitter and which comes from your own analysis.
| Reading object | What it can support | Question to keep visible |
|---|---|---|
| Investigation | Study identity, contacts, design, publications and protocols | What question and workflow does the study describe? |
| Sample sheet | Biological sources, controls and factor values | Which rows define the comparison? |
| Assay file | Sample-to-file links and analytical workflow | Which raw and derived files belong to each sample? |
| Metabolite table | Features, assignments and reported results | What identity evidence and confidence accompany each metabolite? |
Sources and next reading
Substantive resource claims in this article were checked against the primary pages below. A database record is a structured route into evidence; it does not by itself prove a new biological conclusion. Recheck the live record, evidence labels and release context before using it in assessed work or research.
Primary sources
- MetaboLights home — accessed 24 October 2026.
- MetaboLights about — accessed 24 October 2026.
- MetaboLights guides — accessed 24 October 2026.
- MetaboLights files guide — accessed 24 October 2026.
- MetaboLights API — accessed 24 October 2026.