eduKateSG · Why Science?
Bring phenotype, inheritance and variant evidence into one ranked investigation—then open every high-scoring candidate and challenge the reasons it rose
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. Exomiser is an open-source tool that starts with sequence variants and patient phenotypes encoded with the Human Phenotype Ontology. Its current documentation and repository describe annotation, filtering and prioritisation using population frequency, predicted pathogenicity, quality, inheritance patterns, human disease knowledge and cross-species phenotype evidence. The ranked output is a decision-support layer: it does not prove pathogenicity, diagnose a person or replace laboratory validation and expert review. 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: Human Phenotype Ontology and structured clinical phenotypes; Genomics England PanelApp and evidence-rated gene panels; ClinVar variant classifications and evidence conflicts; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 12 October 2026: Official Exomiser repository; Exomiser documentation, release 15.1.1 observed 12 October 2026; Next-generation diagnostics and disease-gene discovery with Exomiser; Phenotype-driven approaches to rare-disease variant prioritisation; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Resource interfaces and recommendations can change, so preserve the release, access date and documentation used.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Prepare | Which case, assembly, pedigree, HPO terms and variant set are being analysed? | Phenopacket or HPO list, VCF, pedigree and release manifest | A versioned input bundle |
| Score | How do variant and phenotype evidence support each candidate? | Frequency, consequence, pathogenicity, inheritance and phenotype similarity | Component scores with provenance |
| Rank | Which variants and genes deserve review first? | Configured filters, combined score and full retained set | A prioritised candidate list |
| Validate | Does independent evidence support the proposed explanation? | Segregation, assay, literature, curation and expert interpretation | A bounded conclusion, not an algorithmic diagnosis |
Inside this guide
1–12 · Foundations and design
- 1. Start with the analysis question
- 2. Understand what Exomiser prioritises
- 3. Encode phenotypes with HPO
- 4. Prepare the VCF carefully
- 5. Add the pedigree and affected status
- 6. Choose inheritance modes deliberately
- 7. Use population frequency as evidence
- 8. Read predicted pathogenicity cautiously
- 9. Read variant consequence in transcript context
- 10. Read quality before rank
- 11. Understand phenotype similarity
- 12. Use model-organism evidence responsibly
13–24 · Models and estimation
- 13. Combine variant and phenotype evidence
- 14. Keep known and novel-gene tasks distinct
- 15. Use phenotype-only mode for the right question
- 16. Treat missing phenotypes as unknown
- 17. Test phenotype specificity
- 18. Test age and onset context
- 19. Inspect retained and filtered variants
- 20. Pin software and data releases
- 21. Read the configuration as methods
- 22. Challenge a top-ranked candidate
- 23. Check variant representation
- 24. Check segregation independently
25–36 · Diagnostics and interpretation
- 25. Check ClinVar and gene–disease evidence separately
- 26. Avoid benchmark leakage
- 27. Protect sensitive data
- 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. Start with the analysis question
Exomiser is most useful when the phenotype, inheritance hypothesis and sequencing result are treated as one evidence problem rather than three disconnected lists.
Working checkpoint — phenotype-driven ranking: Write the case-level question before configuring filters or scores. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 2 of 36
2. Understand what Exomiser prioritises
The tool annotates, filters and ranks likely causative variants and genes; it does not turn the highest row into a confirmed diagnosis.
Working checkpoint — variant frequency: Name the intended use as prioritisation. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 3 of 36
3. Encode phenotypes with HPO
Structured HPO terms let the software compare a case with phenotypes linked to human disease genes and model organisms.
Working checkpoint — inheritance mode: Preserve the original clinical wording beside every mapped term. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 4 of 36
4. Prepare the VCF carefully
A VCF supplies candidate alleles, but calling, normalisation, genome assembly and sample identity determine what the program can see.
Working checkpoint — phenotype similarity: Validate sample names, reference alleles and assembly before analysis. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 5 of 36
5. Add the pedigree and affected status
Family structure and affected status help Exomiser evaluate inheritance and segregation, while errors can redirect the entire ranking.
Working checkpoint — model-organism evidence: Confirm relationships and affected labels with the approved case record. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 6 of 36
6. Choose inheritance modes deliberately
Dominant, recessive, X-linked and other modes imply different allele configurations and frequency expectations.
Working checkpoint — candidate variant: State why each tested mode is biologically plausible. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 7 of 36
7. Use population frequency as evidence
Population data help deprioritise variants too common for some rare high-penetrance models, but frequency depends on ancestry, coverage and dataset release.
Working checkpoint — phenotype-driven ranking: Inspect allele count, number and population context. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 8 of 36
8. Read predicted pathogenicity cautiously
Computational predictors estimate possible molecular impact and can share training data or assumptions.
Working checkpoint — variant frequency: Keep individual predictor outputs and calibration limits visible. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 9 of 36
9. Read variant consequence in transcript context
Exomiser uses transcript definitions through Jannovar, so consequence labels can change with transcript source and genome build.
Working checkpoint — inheritance mode: Report the decisive transcript and annotation release. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 10 of 36
10. Read quality before rank
A plausible biological story cannot rescue a low-quality or misrepresented call without independent confirmation.
Working checkpoint — phenotype similarity: Review depth, balance, mapping and caller flags for leading variants. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 11 of 36
11. Understand phenotype similarity
A similarity score compares the patient profile with annotated disease or gene phenotypes under an ontology and algorithm.
Working checkpoint — model-organism evidence: Open the terms driving a high score. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 12 of 36
12. Use model-organism evidence responsibly
Cross-species comparison can connect human features with mouse, fish and other experimental phenotypes when human evidence is sparse.
Working checkpoint — candidate variant: Translate homologues and phenotype mappings without claiming species identity. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 13 of 36
13. Combine variant and phenotype evidence
The final ranking integrates several signals, so two candidates can reach similar totals by very different evidence routes.
Working checkpoint — phenotype-driven ranking: Inspect components instead of reporting only the combined score. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 14 of 36
14. Keep known and novel-gene tasks distinct
Known-disease diagnosis and novel gene discovery have different evidence expectations and validation burdens.
Working checkpoint — variant frequency: Label which task the configuration is intended to support. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 15 of 36
15. Use phenotype-only mode for the right question
The software can prioritise genes from phenotypes without a VCF, but that output cannot evaluate a patient variant or its segregation.
Working checkpoint — inheritance mode: Call the result a phenotype-ranked gene list. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 16 of 36
16. Treat missing phenotypes as unknown
An unrecorded feature may be unassessed, age-dependent or absent; those states are not interchangeable.
Working checkpoint — phenotype similarity: Keep present, absent, unknown and not-yet-assessed separate. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 17 of 36
17. Test phenotype specificity
Very broad terms match many disorders, while an over-specific or incorrect term can distort the ranking.
Working checkpoint — model-organism evidence: Repeat a planned sensitivity check with defensible term sets. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 18 of 36
18. Test age and onset context
Some phenotypes appear only at certain ages, and onset can distinguish otherwise similar disorders.
Working checkpoint — candidate variant: Record onset and observation age where supported. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 19 of 36
19. Inspect retained and filtered variants
A filtered-out causal variant can never recover through phenotype matching, so the discarded set is scientifically important.
Working checkpoint — phenotype-driven ranking: Save filter reasons for every excluded candidate. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 20 of 36
20. Pin software and data releases
Exomiser software, phenotype data, transcript sources and frequency resources change on different schedules.
Working checkpoint — variant frequency: Archive the application version, data version and configuration. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 21 of 36
21. Read the configuration as methods
Thresholds and prioritiser choices define the analysis and must not remain hidden in an unreported YAML file.
Working checkpoint — inheritance mode: Publish or archive the complete configuration. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 22 of 36
22. Challenge a top-ranked candidate
A high rank can result from incomplete knowledge, a broad phenotype match or correlated predictors.
Working checkpoint — phenotype similarity: Write the strongest competing explanation beside each leader. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 23 of 36
23. Check variant representation
Equivalent alleles can be represented differently after normalisation, left alignment or multiallelic splitting.
Working checkpoint — model-organism evidence: Normalise consistently and retain the source representation. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 24 of 36
24. Check segregation independently
A compatible inheritance score needs confirmation from sample identity, read evidence and appropriate family testing.
Working checkpoint — candidate variant: Review the family genotypes outside the ranking table. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 25 of 36
25. Check ClinVar and gene–disease evidence separately
External classifications and curated relationships add context but can conflict, age or apply to a different allele or disease mechanism.
Working checkpoint — phenotype-driven ranking: Open the exact record and evidence date. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 26 of 36
26. Avoid benchmark leakage
Performance can look better when evaluation cases or associations informed the underlying knowledge resources.
Working checkpoint — variant frequency: Separate historical knowledge cut-offs from test cases. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 27 of 36
27. Protect sensitive data
Phenotype profiles, pedigrees and rare variants can be identifying even without names.
Working checkpoint — inheritance mode: Use approved systems and minimise exported fields. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 28 of 36
28. Connect to Primary Science
A fictional detective activity can rank explanations from clues while showing that the first-ranked answer still needs a test.
Working checkpoint — phenotype similarity: Use invented traits and no personal health information. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 29 of 36
29. Connect to Secondary Science
Genetics, inheritance, mutation and phenotype become a connected evidence chain rather than isolated syllabus terms.
Working checkpoint — model-organism evidence: Label observation, prediction and mechanism in different columns. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 30 of 36
30. Connect to mathematics
Scores, ranks, thresholds and conditional probabilities show why a number needs a scale, denominator and comparison set.
Working checkpoint — candidate variant: Re-rank a toy dataset after changing one justified assumption. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 31 of 36
31. Connect to computing
File formats, identifiers, ontologies, configuration and software versions are part of the scientific argument.
Working checkpoint — phenotype-driven ranking: Add assertions that stop the workflow on sample or assembly mismatch. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Section 32 of 36
32. Connect to pathways
Phenotype-driven genomics links biology, statistics, medicine, software and data stewardship and can illuminate STEM pathways.
Working checkpoint — variant frequency: Check current official study and professional requirements separately. Read the combined rank beside its frequency, consequence, pathogenicity, inheritance and phenotype-similarity components.
For future study, connect genomics to biology, statistics, software and ethics while checking official pathways separately.
Section 33 of 36
33. Did you know?
Exomiser can use phenotypes from model organisms to help prioritise human candidate genes when direct human disease annotation is limited.
Working checkpoint — inheritance mode: Treat cross-species similarity as supportive evidence, not proof. Preserve filtered variants and rejection reasons; a top list is not the complete analysis.
For scientific writing, name the Exomiser and data versions whenever a rank is reported.
Section 34 of 36
34. Write a reader-ready result
A clear report names the case inputs, releases, tested modes, filters, component evidence, leading candidates and major uncertainties.
Working checkpoint — phenotype similarity: Explain why each candidate ranked highly. Separate what Exomiser prioritises from what laboratory and expert review later establish.
For students, ask which clue came from phenotype, inheritance, population frequency or molecular prediction.
Section 35 of 36
35. Build a validation ladder
Verify inputs and representation, inspect component scores, confirm segregation and literature, then choose orthogonal laboratory or clinical review.
Working checkpoint — model-organism evidence: Strengthen the weakest evidence rung first. Treat sample names, assemblies, transcripts and identifiers as scientific data: a silent mismatch can redirect the ranking.
For parents, ask whether the result is a prioritised candidate or an independently validated finding.
Section 36 of 36
36. Finish with a reproducible bundle
Deliver permitted inputs, checksums, configuration, software and data versions, complete outputs, plots, review notes and validation status.
Working checkpoint — candidate variant: Make the ranking reconstructable without exposing unnecessary identifiers. Keep the HPO profile, VCF, pedigree, assembly, configuration, software and data versions together.
For a classroom model, use fictional creatures, alleles and pedigrees rather than real health data.
Contents · Previous section · Continue to the Science Learning Hub