eduKateSG · Why Science?
Trace every variant–trait association through its study, effect allele, population descriptors, statistics and mapped trait before asking what it means
Choose the closest route, then return to the full index whenever you need the wider evidence chain.
Science learning becomes powerful when students can tell a measured pattern from a model, a statistical decision from a mechanism, and a useful lead from a finished conclusion. The NHGRI-EBI GWAS Catalog is a curated, searchable resource for genome-wide association studies, reported variant–trait associations, study metadata and deposited summary statistics. Its official documentation says eligible studies are identified and curated, with reported traits, significant associations and sample metadata extracted into structured records; literature-curated top associations and full summary-statistics deposits are different data layers. A catalogued association means an allele and trait covaried statistically in a stated study. It does not by itself identify the causal variant, causal gene, biological mechanism or individual outcome. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science and STEM education as a longform evidence-reading route; it is not medical advice, a laboratory protocol, an admission promise or a career guarantee.
Related eduKate reading: gnomAD and population allele-frequency context; Ensembl VEP and transcript-aware variant consequences; ANNOVAR and gene-, region- and filter-based annotation; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official NHGRI-EBI GWAS Catalog about page; Official GWAS Catalog curation methods; Official GWAS Catalog REST API documentation; GWAS Catalog knowledgebase paper in PubMed; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Package interfaces and recommendations can change, so reproduce an analysis only with the documentation for the installed release.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Study | What was measured, in whom, and under which design? | Publication, cohort, phenotype and population descriptors | A scoped study record |
| Association | Which effect allele was linked to which trait and statistic? | P-value, effect size, confidence interval and risk frequency | A reproducible reported association |
| Context | Could linkage, mapping or design explain the signal? | LD structure, locus annotation, replication and summary statistics | A set of causal hypotheses |
| Validate | What independent evidence supports mechanism or prediction? | Fine mapping, experiments, diverse cohorts and prospective validation | A bounded conclusion beyond association |
Inside this guide
1–12 · Foundations and design
- 1. Begin with association, not causation
- 2. Understand the Catalog’s job
- 3. Read eligibility and curation rules
- 4. Keep the reported trait and mapped trait
- 5. Read the P-value correctly
- 6. Respect genome-wide thresholds
- 7. Identify the effect allele
- 8. Read effect size and uncertainty
- 9. Separate discovery from replication
- 10. Distinguish top associations from summary statistics
- 11. Account for multiple testing
- 12. Understand linkage disequilibrium
13–24 · Models and estimation
- 13. Do not assign the nearest gene automatically
- 14. Freeze assembly and dbSNP context
- 15. Separate allele frequency from effect
- 16. Distinguish binary and quantitative traits
- 17. Check sample overlap
- 18. Read population descriptors
- 19. Challenge transferability
- 20. Expect publication and selection effects
- 21. Use ontology searches deliberately
- 22. Use the REST API for focused questions
- 23. Use flat files for bulk access
- 24. Track the access date
25–36 · Diagnostics and interpretation
- 25. Reanalyse rather than overread a row
- 26. Keep prediction separate
- 27. Use bounded clinical language
- 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 result
- 35. Build a validation ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Begin with association, not causation
GWAS tests whether genetic variants correlate with traits across studied samples. A significant result is a statistical relationship under that design.
Working checkpoint — genome-wide association: Use “associated with” until causal evidence is available. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 2 of 36
2. Understand the Catalog’s job
The resource curates eligible studies, reported associations, traits and sample metadata into searchable records. It is a knowledgebase and deposition resource, not one giant experiment.
Working checkpoint — effect allele: Cite the study accession and publication for every claim. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 3 of 36
3. Read eligibility and curation rules
Catalog inclusion follows documented study and association criteria. Curators standardise fields without erasing the design decisions of the source paper.
Working checkpoint — P-value: Read the methods page before treating the collection as exhaustive. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 4 of 36
4. Keep the reported trait and mapped trait
Authors’ trait language may differ across papers. The Catalog maps traits to ontology terms so related studies can be searched consistently.
Working checkpoint — effect size: Retain both the reported trait and mapped ontology identifier. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 5 of 36
5. Read the P-value correctly
A P-value measures compatibility with a null model under assumptions; it is not the probability that the association is true or important.
Working checkpoint — linkage disequilibrium: Report the exact value and multiple-testing context. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 6 of 36
6. Respect genome-wide thresholds
Very small thresholds control the many tests in a GWAS. The Catalog’s diagram highlights genome-wide-significant associations, while curation and downloads may include broader reported data.
Working checkpoint — summary statistics: State the inclusion threshold used for the exact dataset. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 7 of 36
7. Identify the effect allele
The association direction belongs to a specific effect or risk allele. Flipping alleles without flipping the effect creates a false conclusion.
Working checkpoint — genome-wide association: Store effect allele, other allele, strand and harmonisation status. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 8 of 36
8. Read effect size and uncertainty
Odds ratios and beta coefficients quantify different outcome models. Confidence intervals show uncertainty that a P-value alone hides.
Working checkpoint — effect allele: Report effect measure, units and interval together. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 9 of 36
9. Separate discovery from replication
Discovery and replication stages can use different samples and analyses. A replicated association is stronger than an unconfirmed discovery signal, but still not necessarily causal.
Working checkpoint — P-value: Record which cohort supports each stage. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 10 of 36
10. Distinguish top associations from summary statistics
Literature-curated top associations capture selected reported signals. Full summary statistics provide results across the genome for eligible deposits.
Working checkpoint — effect size: Choose the data layer that matches the question. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 11 of 36
11. Account for multiple testing
Millions of correlated tests create false positives if ordinary thresholds are used. Study-specific correction and quality control define the evidence.
Working checkpoint — linkage disequilibrium: Keep the analysis threshold and number of tests visible. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 12 of 36
12. Understand linkage disequilibrium
Nearby variants can be inherited together, so the strongest associated marker may tag another causal allele. Association localises a region more readily than a mechanism.
Working checkpoint — summary statistics: Inspect LD and fine-mapping evidence before naming a causal variant. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 13 of 36
13. Do not assign the nearest gene automatically
A lead variant can regulate a distant gene, lie in a shared region or tag another allele. Nearest-gene annotation is a hypothesis.
Working checkpoint — genome-wide association: Separate mapped genes, reported genes and experimentally supported targets. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 14 of 36
14. Freeze assembly and dbSNP context
Coordinates and identifiers are mapped to specific genome and dbSNP builds. Release changes can move or merge records.
Working checkpoint — effect allele: Record Catalog release date, assembly and variant identifier source. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 15 of 36
15. Separate allele frequency from effect
A common allele can have a small effect, while a rare allele can have a larger one. Frequency and effect size answer different questions.
Working checkpoint — P-value: Report both without using one as a substitute for the other. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 16 of 36
16. Distinguish binary and quantitative traits
Odds ratios often describe case-control outcomes; beta values may describe continuous measurements. Units and transformation shape interpretation.
Working checkpoint — effect size: Name the phenotype scale and model. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 17 of 36
17. Check sample overlap
Two studies can share participants or controls, making their evidence less independent than publication count suggests. Meta-analysis must account for overlap.
Working checkpoint — linkage disequilibrium: Review cohort names and recruitment sources. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 18 of 36
18. Read population descriptors
Ancestry, country of recruitment and sample descriptors influence allele frequencies, LD and transferability. Labels require careful reading and should not become stereotypes.
Working checkpoint — summary statistics: Use the Catalog’s standardised descriptors with the original study context. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 19 of 36
19. Challenge transferability
An association estimated in one sampled population may not predict equally in another because of LD, environment, frequency and model differences.
Working checkpoint — genome-wide association: Seek diverse replication and recalibration. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 20 of 36
20. Expect publication and selection effects
Published top hits are a selected subset of all analyses. Negative, underpowered or unpublished work may be less visible.
Working checkpoint — effect allele: Use summary-statistics deposits and preregistered analyses where appropriate. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 21 of 36
21. Use ontology searches deliberately
Including child traits can broaden a query substantially. The REST API documentation makes this choice explicit for trait endpoints.
Working checkpoint — P-value: Save the ontology ID and child-trait setting. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 22 of 36
22. Use the REST API for focused questions
The API can search literature-curated associations, traits, population descriptors and studies with summary statistics. It is not the preferred path for bulk downloading everything.
Working checkpoint — effect size: Record endpoint, parameters, date and response schema. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 23 of 36
23. Use flat files for bulk access
The official API documentation recommends flat files for bulk retrieval rather than crawling the service. Bulk files preserve a release-like snapshot.
Working checkpoint — linkage disequilibrium: Checksum downloads and archive the data dictionary. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 24 of 36
24. Track the access date
The Catalog is updated regularly and the about page describes a bi-weekly release cycle. Counts and records can change after curation or deposition.
Working checkpoint — summary statistics: Cite the access date and data release. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 25 of 36
25. Reanalyse rather than overread a row
A Catalog entry is a starting point. Summary statistics, harmonisation, LD reference, phenotype definition and model assumptions determine downstream results.
Working checkpoint — genome-wide association: Reproduce a small locus before scaling. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 26 of 36
26. Keep prediction separate
A GWAS association can inform a risk model only after suitable training, validation, calibration and governance. One odds ratio is not an individual forecast.
Working checkpoint — effect allele: Do not turn Catalog rows into health advice. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 27 of 36
27. Use bounded clinical language
The Catalog supports research discovery and evidence lookup. It does not diagnose disease or establish treatment for a person.
Working checkpoint — P-value: Keep educational explanations population-level and non-directive. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 28 of 36
28. Connect to Primary Science
A class survey can show correlation between two features while a hidden third factor explains both.
Working checkpoint — effect size: Use harmless invented data and ask what was not measured. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 29 of 36
29. Connect to Secondary Science
Inheritance, variation and phenotype meet statistical design. Students can see why a marker may travel with, rather than cause, an effect.
Working checkpoint — linkage disequilibrium: Draw a locus with linked variants before discussing a gene. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 30 of 36
30. Connect to mathematics
Effect sizes, confidence intervals, multiple testing and sampling turn association into a quantitative claim.
Working checkpoint — summary statistics: Read the interval and denominator before the headline. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 31 of 36
31. Connect to computing
Ontologies, stable study accessions, APIs, flat files and summary-statistics schemas make thousands of studies interoperable.
Working checkpoint — genome-wide association: Version the query and validate allele orientation. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 32 of 36
32. Connect to pathways
Statistical genetics joins biology, mathematics, epidemiology and data engineering and can illuminate STEM education and career pathways.
Working checkpoint — effect allele: Check current official programmes for real pathway decisions. Read the numerical result together with sample-level plots, effect size and a sensitivity check capable of weakening the headline.
For future study, connect the idea to biology, mathematics and computing while checking current official pathway information separately.
Section 33 of 36
33. Did you know?
The Catalog distinguishes curated top associations from deposited full summary statistics. A headline row and a genome-wide result file answer different questions.
Working checkpoint — P-value: Choose the correct layer before analysis. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 34 of 36
34. Write a reader-ready result
A strong summary names study, trait, effect allele, statistic, effect size, confidence interval, sample descriptors, assembly and whether the signal was replicated.
Working checkpoint — effect size: End with “association” unless mechanism has independent support. Separate what the model estimates from the biological story proposed afterward, then name the extra evidence that story would require.
For students, the transferable habit is to ask: What was measured? Compared with what? What alternative explanation remains?
Section 35 of 36
35. Build a validation ladder
Verify the row, inspect the study, check allele orientation, examine LD and summary statistics, replicate across cohorts, then test biological mechanism.
Working checkpoint — linkage disequilibrium: Do not skip from P-value to causal gene. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 36 of 36
36. Finish with a reproducible bundle
Deliver study and association accessions, publication, exact query, ontology IDs, release files, summary-statistics source, harmonisation code, LD reference and validation record.
Working checkpoint — summary statistics: Make every association claim reconstructable. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Contents · Previous section · Continue to the Science Learning Hub