eduKateSG · Why Science?
Describe who ate what, when and under which study design—then keep measurements and limitations attached
Start with the nutrition-study question, then trace participants, exposure or intervention, food identity, measurements, time and analysis before comparing evidence.
Science learning becomes powerful when a nutrition claim can be traced to a clearly described study rather than a headline about one food. The OBO Foundry says the Ontology for Nutritional Studies, or ONS, was developed through the ENPADASI European project to support standardised descriptions of human nutritional studies. Its official page says ONS covers concepts and relations used when conducting, storing, harmonising, integrating, describing and querying intervention and observational studies, and lists dependencies including BFO, ChEBI, ENVO, FoodOn, NCBI Taxonomy, OBI, RO and Uberon. ONS can organise study meaning, but it cannot repair biased recruitment, inaccurate dietary recall, poor adherence, confounding or an unjustified causal claim. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science, O-Level Biology, food science, mathematics, computing and STEM education as a longform evidence-reading route; it is not dietary, medical, food-safety, admission or career advice.
Related eduKate reading: FoodOn and interoperable food data; Statistics Ontology; ChEBI and stable chemical meaning; Ontology for Biomedical Investigations; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 13 October 2026: OBO Foundry Ontology for Nutritional Studies overview, checked 13 October 2026; Official ONS source repository; Canonical ONS OWL release; ONS publication listed by OBO Foundry; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Ontologies, definitions and release products can change, so preserve dated documentation, exact source records and file versions.
How a nutritional study becomes reusable scientific evidence
Nutrition evidence links people, foods, nutrients, behaviours, measurements and time. A study description that captures only the food name cannot explain who consumed it, in what dose, under what design or with which outcome.
The OBO Foundry describes ONS as supporting both interventions and observational studies. Those designs answer different questions: assigning an intervention and observing habitual diet involve different bias and causal assumptions.
ONS was designed for conducting, storing, harmonising, integrating, describing and querying studies. Reuse therefore depends on operational details, not simply a publication citation or abstract.
The official dependency list connects ONS with FoodOn for foods, ChEBI for chemicals, ENVO for environment, OBI for investigations, NCBI Taxonomy for organisms and Uberon for anatomy. Integration requires frozen versions of those dependencies.
Food identity needs processing and preparation context. Raw, boiled, fermented or fortified versions of a food can differ materially while sharing an everyday name.
Dietary exposure can be assigned, reported, purchased, photographed or inferred from biomarkers. Each method measures a different layer and carries characteristic error.
Nutrition outcomes span physiology, behaviour and disease-related measures. Timing matters because a short post-meal response and a long-term health outcome are not interchangeable.
The happy purpose is cumulative learning: researchers can compare study designs and measurements while readers see exactly where evidence is strong, uncertain or unable to establish causation.
Begin with the target population and eligibility criteria. Age, health status, location and baseline diet influence whether findings apply beyond the enrolled sample.
Record intervention dose, frequency, duration, comparator and co-interventions. A named diet can otherwise conceal substantial variation between studies.
Adherence is evidence, not an assumption. Attendance, diaries, returned products and biomarkers may disagree, and the method used should be visible.
Dietary recall is vulnerable to memory, estimation and social-desirability error. Ontology can describe the method but cannot make the report objectively complete.
Energy intake and substitution matter. Adding one food may displace another, so an observed effect cannot always be assigned to the added item alone.
Outcome methods need units, calibration and specimen handling. A laboratory analyte, questionnaire scale and clinical event require different validation.
For classroom learning, students can compare two breakfast surveys and discover how portion definition, recall period and participant group change the meaning of a graph.
Reader-ready reporting separates what participants were asked to do, what they actually did, what was measured and what the analysis estimated.
Confounding is central in observational nutrition research because diet travels with income, culture, activity, smoking, sleep and access to care. Adjustment reduces some differences only under stated assumptions.
Randomisation strengthens causal inference for assigned interventions, yet attrition, non-adherence, blinding limits and measurement error can still weaken conclusions.
Composite dietary scores compress many choices into one number. Reuse requires component definitions, scoring rules, missing-data handling and the version of the index.
Nutrient databases and recipes change over time. Preserve the food-composition source, edition, matching decisions and yield or retention factors behind derived nutrient intake.
Multiple outcomes and subgroup analyses create many chances for striking results. Distinguish planned, exploratory and sensitivity analyses.
Statistical significance does not establish practical importance, individual benefit or safety. Report effect size, uncertainty, baseline risk and study limitations.
A reproducible study bundle includes protocol, dictionaries, ontology releases, food mappings, code, measurement metadata, analysis plan, diagnostics and amendment history.
ONS helps organise the study story; responsible science still asks whether the design, participants, exposure measurement and analysis deserve the conclusion.
Nutrition studies also need a clear estimand. Researchers may ask about the effect of assignment to a diet, the effect among people who adhered, the association between reported intake and an outcome, or the consequence of replacing one food with another. Those targets are not interchangeable. Record the population, contrast, time horizon and handling of intercurrent events such as treatment changes or dropout. The analysis model and reported conclusion should point to the same estimand. Otherwise a well-structured study graph can still connect an exposure, outcome and result whose statistical meaning does not match the question readers think was asked.
Cultural and regional food practices are scientific context, not noise. Recipes, portion sizes, meal patterns and food names vary across communities, and a standard food concept may need local preparation detail to represent actual exposure. Translation should preserve the source term and involve reviewers who understand the cuisine. When a composition database lacks an exact match, record the substitute and uncertainty rather than choosing the nearest familiar item invisibly. This makes cross-country research more honest and prevents harmonisation from systematically reducing the diets of underrepresented populations to inaccurate approximations.
Safety and benefit cannot be inferred from nutrient identity alone. Dose, baseline status, interactions, vulnerable groups and duration can change the balance of effects. An ontology can connect a nutrient, supplement and outcome measure, but clinical interpretation still requires current evidence and professional standards. Public-facing writing should avoid turning a study association into an instruction for an individual reader. State the study population, absolute changes, uncertainty and adverse-event reporting, and direct personal health decisions to qualified professionals. This boundary protects both scientific usefulness and the people who may act on a simplified headline.
Open and FAIR-style study descriptions work only when rights and privacy are respected. Participant-level diet, location and health records can be identifying, particularly when rare conditions or small communities are involved. Publish metadata and aggregate evidence at a level compatible with consent, governance and disclosure risk. Controlled access may be appropriate for detailed records, while code, data dictionaries, ontology mappings and synthetic fixtures can remain open. Reusability is therefore not maximum exposure; it is the ability to find, understand and responsibly request or reproduce what the study can lawfully share.
A classroom mini-study can model these principles safely using anonymous, low-stakes observations such as labelled breakfast categories from fictional characters. Students define categories, decide how to record portion uncertainty, build a small graph and test whether two groups are comparable. They then rewrite a sensational conclusion into a calibrated one. This connects Primary and Secondary Science, mathematics, computing and English: the data structure preserves who, what and when; the calculations show variation; and the writing distinguishes observation from inference without offering personal dietary advice.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Frame | What nutrition question and population are studied? | Protocol, eligibility, design and time frame | A bounded study question |
| Describe | What food, nutrient or intervention occurred? | ONS, FoodOn, ChEBI, dose, schedule and adherence | A traceable exposure record |
| Measure | Which outcomes and covariates were collected? | Method, unit, specimen, timing and uncertainty | Comparable observations |
| Interpret | What conclusion does the design support? | Analysis, confounding, limitations and provenance | A calibrated evidence statement |
Inside this guide
1–12 · Study question, people and exposure
- 1. Begin with the nutrition question
- 2. Identify the study design
- 3. Use the ONS identifier
- 4. Define the population
- 5. Define the intervention
- 6. Define observational exposure
- 7. Identify the food precisely
- 8. Identify chemical components
- 9. Record preparation and processing
- 10. Record dose and frequency
- 11. Record duration and timing
- 12. Measure adherence
13–24 · Measurements, analysis and comparison
- 13. Describe the comparator
- 14. Describe outcomes
- 15. Describe specimens
- 16. Describe measurement units
- 17. Describe covariates
- 18. Handle dietary recall
- 19. Handle food composition mapping
- 20. Handle missing data
- 21. Pre-specify the analysis
- 22. Separate planned and exploratory work
- 23. Challenge confounding
- 24. Challenge substitution blindness
25–36 · Learning, communication and pathways
- 25. Challenge adherence assumptions
- 26. Challenge headline causality
- 27. Review ontology and database migrations
- 28. Connect to Primary Science
- 29. Connect to Secondary Science
- 30. Connect to mathematics
- 31. Connect to computing
- 32. Connect to pathways
- 33. Did you know?
- 34. Write a reader-ready study note
- 35. Build a validation ladder
- 36. Finish with a reproducible nutrition bundle
Section 1 of 36
1. Begin with the nutrition question
A reusable study starts with population, exposure or intervention, comparator, outcome and time rather than a fashionable food name.
Working checkpoint — ONS identifier: Write the complete question before choosing variables. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 2 of 36
2. Identify the study design
Randomised interventions, cohorts, cross-sectional surveys and feeding studies create different evidence and bias structures.
Working checkpoint — nutritional intervention: Name assignment, follow-up and comparison plan. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 3 of 36
3. Use the ONS identifier
ONS classes give study concepts stable references beyond local forms and abbreviations.
Working checkpoint — observational study: Store identifier, label, definition and release. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 4 of 36
4. Define the population
Eligibility, recruitment setting and baseline health shape who the evidence represents.
Working checkpoint — dietary exposure: Record inclusion, exclusion and sampling frame. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 5 of 36
5. Define the intervention
A diet or supplement needs dose, schedule, duration, delivery and comparator.
Working checkpoint — outcome measurement: Model each component and change over time. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 6 of 36
6. Define observational exposure
Habitual intake may be self-reported, observed, purchased, photographed or inferred.
Working checkpoint — study provenance: State the measurement layer and recall period. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 7 of 36
7. Identify the food precisely
FoodOn can distinguish source, processing and product identity that everyday labels collapse.
Working checkpoint — ONS identifier: Map the source item and retain original wording. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 8 of 36
8. Identify chemical components
ChEBI identifiers can represent nutrients and other chemicals when the measured entity is known.
Working checkpoint — nutritional intervention: Keep substance, amount, basis and analytical method together. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 9 of 36
9. Record preparation and processing
Cooking, fermentation, fortification and storage can change composition and exposure.
Working checkpoint — observational study: Describe the consumed form, not only the raw ingredient. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 10 of 36
10. Record dose and frequency
One serving once and repeated daily exposure answer different biological questions.
Working checkpoint — dietary exposure: Store amount, unit, schedule and permitted range. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 11 of 36
11. Record duration and timing
Acute post-meal responses differ from long-term changes in health indicators.
Working checkpoint — outcome measurement: Anchor intervention and outcome times to the protocol. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 12 of 36
12. Measure adherence
The assigned plan and actual behaviour can diverge throughout a study.
Working checkpoint — study provenance: Record adherence method, observations and uncertainty. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 13 of 36
13. Describe the comparator
Usual diet, placebo, alternative food and baseline period lead to different contrasts.
Working checkpoint — ONS identifier: Name what changed and what was held constant. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 14 of 36
14. Describe outcomes
Anthropometry, biomarkers, symptoms and clinical events require distinct methods and interpretation.
Working checkpoint — nutritional intervention: Use stable outcome identities and operational definitions. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 15 of 36
15. Describe specimens
Blood, urine, tissue and other samples need collection, processing, storage and anatomical context.
Working checkpoint — observational study: Link specimen records to participant and time point. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 16 of 36
16. Describe measurement units
Concentration, mass, energy and frequency values need units and reference bases.
Working checkpoint — dietary exposure: Validate dimensions and preserve source precision. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 17 of 36
17. Describe covariates
Age, activity, medication, smoking and socioeconomic factors may affect exposure and outcome.
Working checkpoint — outcome measurement: State measurement and modelling role for each covariate. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 18 of 36
18. Handle dietary recall
Memory and portion estimation can produce systematic as well as random error.
Working checkpoint — study provenance: Publish instruments, coding rules and quality checks. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 19 of 36
19. Handle food composition mapping
Reported foods must be matched to a nutrient database or recipe model.
Working checkpoint — ONS identifier: Record database edition, match and conversion factors. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 20 of 36
20. Handle missing data
Missing visits, diaries or specimens can change the analysed population.
Working checkpoint — nutritional intervention: Report patterns, reasons and handling assumptions. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 21 of 36
21. Pre-specify the analysis
A plan written before outcome inspection reduces outcome switching and hidden flexibility.
Working checkpoint — observational study: Version the protocol, estimands and statistical methods. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 22 of 36
22. Separate planned and exploratory work
Exploration can generate hypotheses but should not inherit the certainty of a prespecified test.
Working checkpoint — dietary exposure: Label analysis status and amendments. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 23 of 36
23. Challenge confounding
Observational associations can reflect lifestyle, access or health differences linked to diet.
Working checkpoint — outcome measurement: State causal assumptions and remaining alternatives. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 24 of 36
24. Challenge substitution blindness
Adding one food often means reducing another, which changes the scientific contrast.
Working checkpoint — study provenance: Model replacement and total intake where relevant. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 25 of 36
25. Challenge adherence assumptions
A supplied product does not prove it was consumed as planned.
Working checkpoint — ONS identifier: Compare adherence sources and run sensitivity analyses. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 26 of 36
26. Challenge headline causality
A narrow association or short experiment rarely supports a universal health promise.
Working checkpoint — nutritional intervention: Match verbs and scope to design and population. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 27 of 36
27. Review ontology and database migrations
Changed classes, food databases or nutrient values can alter derived exposures.
Working checkpoint — observational study: Rerun mappings and quantify affected records. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 28 of 36
28. Connect to Primary Science
Simple food observations can teach classification, fair comparison and the difference between claim and evidence.
Working checkpoint — dietary exposure: Use safe examples without personal dietary advice. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 29 of 36
29. Connect to Secondary Science
Human nutrition, enzymes, transport and homeostasis connect foods to biological systems and experimental design.
Working checkpoint — outcome measurement: Trace one claim from intake measure to outcome. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 30 of 36
30. Connect to mathematics
Rates, averages, variability, energy balance and uncertainty make study conclusions quantitative.
Working checkpoint — study provenance: Keep denominators and participant counts visible. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 31 of 36
31. Connect to computing
Schemas, ontologies and provenance link participants, foods, samples, time points and analyses.
Working checkpoint — ONS identifier: Validate foreign keys, units and release versions. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Section 32 of 36
32. Connect to pathways
Nutrition-study semantics matters in food science, epidemiology, public health, clinical research and data stewardship.
Working checkpoint — nutritional intervention: Check current official course and career information separately. Separate assigned intervention, reported intake, biomarker evidence and inferred nutrient values.
For computing, validate foods, units, time points and participant links.
Section 33 of 36
33. Did you know?
The official ONS page lists FoodOn, ChEBI, ENVO, OBI, NCBI Taxonomy and Uberon among its dependencies.
Working checkpoint — observational study: Follow one study record across those semantic layers. Pin ONS and imported ontology releases before combining studies.
For writing, distinguish intervention, observation, association and causal conclusion.
Section 34 of 36
34. Write a reader-ready study note
Name population, design, exposure, comparator, adherence, outcomes, analysis, effect, uncertainty and limitations.
Working checkpoint — dietary exposure: Let readers judge the evidence without relying on a headline. Report effect size, uncertainty, confounding and limitations beside every nutrition claim.
For students, ask who ate what, how it was measured and what changed.
Section 35 of 36
35. Build a validation ladder
Check study identity, design, participants, foods, dose, timing, measurements, analysis, provenance and interpretation.
Working checkpoint — outcome measurement: Stop confidence at the earliest unresolved rung. Preserve protocol, food mappings, composition source, code and amendment history.
For parents, the calm checkpoint is whether a headline matches the study design.
Section 36 of 36
36. Finish with a reproducible nutrition bundle
Deliver protocol, instruments, dictionaries, ontology releases, food mappings, composition sources, code and limitations.
Working checkpoint — study provenance: Make every conclusion traceable to study records. Keep population, design, exposure, dose, adherence, outcome, timing and analysis together.
For mathematics, compare effect size and uncertainty instead of a threshold alone.
Contents · Previous section · Continue to the Science Learning Hub