eduKateSG · Why Science?
Ask three different annotation questions—gene, region and filter—and keep every database answer versioned
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. ANNOVAR is a software system for functionally annotating genetic variants from diverse genomes. Its documentation separates gene-based annotation, region-based annotation and filter-based annotation: those modes ask different questions and can be combined through table_annovar.pl protocols. The official site also publishes dated updates for resources such as ClinVar, dbNSFP and population-frequency collections. An annotation is only as current and accurate as its assembly, database release, allele representation and protocol settings; it is not a clinical verdict. 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: Ensembl VEP and transcript-aware variant consequences; SnpEff and genome-scale variant impact annotation; MMSplice and modular splice-variant effects; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official ANNOVAR documentation; Official ANNOVAR quick start guide; ANNOVAR primary study full text; ANNOVAR primary study record 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 |
|---|---|---|---|
| Gene | How does the allele relate to a transcript model? | Assembly-matched gene definition | Exonic, splicing or other gene-relative consequence |
| Region | Which genomic intervals overlap it? | BED-like tracks or curated region database | A documented interval overlap |
| Filter | Does the exact allele match a reference collection? | Versioned allele database | Match, frequency or database field |
| Interpret | What claim survives independent review? | Complete protocols plus orthogonal evidence | A bounded, reproducible interpretation |
Inside this guide
1–12 · Foundations and design
- 1. Begin with three annotation modes
- 2. Choose the genome build
- 3. Prepare explicit alleles
- 4. Use conversion tools transparently
- 5. Understand gene-based annotation
- 6. Understand region-based annotation
- 7. Understand filter-based annotation
- 8. Read table_annovar protocols
- 9. Freeze downloaded resources
- 10. Respect current documentation
- 11. Compare transcript definitions
- 12. Keep transcript versions
13–24 · Models and estimation
- 13. Read exonic-function labels carefully
- 14. Read splicing labels carefully
- 15. Interpret intergenic distance
- 16. Use population frequencies with ancestry context
- 17. Use ClinVar as dated evidence
- 18. Use dbNSFP fields independently
- 19. Build custom databases carefully
- 20. Check duplicate and multiple matches
- 21. Handle multiallelic records explicitly
- 22. Handle insertions and deletions carefully
- 23. Track missing database values
- 24. Compare with VEP and SnpEff fairly
25–36 · Diagnostics and interpretation
- 25. Inspect command logs
- 26. Audit one variant end to end
- 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 the result responsibly
- 35. Build a validation ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Begin with three annotation modes
ANNOVAR distinguishes gene-based, region-based and filter-based annotation. Treating them as one generic lookup hides the question each mode answers.
Working checkpoint — gene-based annotation: Label every protocol as gene, region or filter before running it. 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. Choose the genome build
The buildver parameter selects the coordinate system and database prefix. hg19, hg38 and hs1 are not interchangeable maps.
Working checkpoint — region-based annotation: Record build and validate reference alleles against the matching FASTA. 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. Prepare explicit alleles
ANNOVAR input identifies chromosome, start, end, reference and alternative alleles. Insertions and deletions require exact representation.
Working checkpoint — filter-based annotation: Keep the original VCF beside the converted avinput file. 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. Use conversion tools transparently
convert2annovar.pl can transform VCF and other formats into ANNOVAR input. Conversion options determine how multiallelic records, genotypes and extra fields survive.
Working checkpoint — protocol operation: Document the conversion command and compare row counts. 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. Understand gene-based annotation
Gene-based mode asks how a variant relates to transcript structures and coding sequence. Results depend on the selected gene definition, such as refGene or ensGene.
Working checkpoint — database release: Name the gene database and its dated release. 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. Understand region-based annotation
Region-based mode asks whether a variant overlaps a genomic interval such as a conserved region, regulatory track or segmental duplication.
Working checkpoint — independent evidence: Describe the region source, coordinate build and overlap rule. 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. Understand filter-based annotation
Filter-based mode asks whether an allele matches a database record, often to retrieve frequency or evidence fields. Coordinate-only matches can be misleading at multiallelic sites.
Working checkpoint — gene-based annotation: Verify allele-specific matching and database encoding. 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 table_annovar protocols
table_annovar.pl can combine several databases in a declared protocol list with corresponding operation codes. Protocol order and operation type create the output schema.
Working checkpoint — region-based annotation: Save the exact protocol and operation strings. 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. Freeze downloaded resources
The official ANNOVAR page publishes dated resource updates. A ClinVar or dbNSFP label without its date or version cannot be reproduced reliably.
Working checkpoint — filter-based annotation: Record each database filename, keyword, download date and checksum. 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. Respect current documentation
The official site checked on 11 October 2026 listed updates through September 2026, including dbNSFP 5.4a and a dated ClinVar build. Those facts are current snapshots, not permanent defaults.
Working checkpoint — protocol operation: Cite the resource actually used rather than the newest item on the page. 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. Compare transcript definitions
refGene, knownGene and ensGene can assign different transcripts and coding consequences. Disagreement often reflects different models rather than a simple software error.
Working checkpoint — database release: Retain transcript identifiers and compare critical cases visually. 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. Keep transcript versions
A transcript accession without its version can point to a changing sequence model. Small version changes may alter coding position or amino-acid notation.
Working checkpoint — independent evidence: Use versioned accessions where the selected resource provides them. 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. Read exonic-function labels carefully
Labels such as synonymous, nonsynonymous, stopgain or frameshift describe predicted coding consequences under one transcript model. They do not measure protein function.
Working checkpoint — gene-based annotation: Separate coding annotation from protein and phenotype evidence. 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. Read splicing labels carefully
A splicing designation follows configured distance rules around exon boundaries. It predicts positional relevance but does not observe RNA processing.
Working checkpoint — region-based annotation: Record the splicing threshold and seek transcript evidence. 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. Interpret intergenic distance
Nearest-gene outputs provide orientation, not proof that the nearby gene is regulated by the variant. Three-dimensional regulation may skip the nearest gene.
Working checkpoint — filter-based annotation: Use proximity language and add functional regulatory evidence separately. 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. Use population frequencies with ancestry context
Reference frequencies help evaluate rarity only within their sampled populations, quality filters and release. Missingness and representation differ across groups.
Working checkpoint — protocol operation: Report population-specific fields and callability where 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 17 of 36
17. Use ClinVar as dated evidence
ClinVar records can change as submitters add evidence or classifications are reviewed. A downloaded snapshot is not a timeless truth table.
Working checkpoint — database release: Keep assertion, review status, condition and database date. 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. Use dbNSFP fields independently
dbNSFP aggregates many missense and functional scores. Correlated predictors and differing transcripts mean columns should not be treated as independent votes.
Working checkpoint — independent evidence: Document score version, transcript mapping and missing-value rule. 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. Build custom databases carefully
ANNOVAR supports user-built filter or region databases. Formatting and sorting determine whether queries return correct matches.
Working checkpoint — gene-based annotation: Validate custom resources with positive and negative test variants. 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. Check duplicate and multiple matches
A variant may match several transcripts, clinical records or frequency entries. Flattening them to one cell can hide conflicts.
Working checkpoint — region-based annotation: Preserve delimiters and expand critical records during review. 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. Handle multiallelic records explicitly
Different alternative alleles at one coordinate can have different frequencies and consequences. Position-only merging risks transferring evidence between alleles.
Working checkpoint — filter-based annotation: Split, normalise and test allele-aware joins. 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. Handle insertions and deletions carefully
Indels can have equivalent left- or right-shifted representations and may span coding boundaries. Database matching can fail even when biology is equivalent.
Working checkpoint — protocol operation: Normalise against the same reference and inspect unmatched important alleles. 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. Track missing database values
A dot or blank can mean the allele is absent, the region was not covered, the file failed to load or the key did not match. Those meanings are not equivalent.
Working checkpoint — database release: Assign explicit missingness categories before filtering. 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. Compare with VEP and SnpEff fairly
Different tools use different annotations, vocabularies and summary rules. Agreement becomes informative only after inputs and transcript sources are aligned.
Working checkpoint — independent evidence: Compare feature-level rows rather than headline severity alone. 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. Inspect command logs
A completed output file does not prove every database loaded successfully. Warnings, HTTP failures and column shifts can produce partial tables.
Working checkpoint — gene-based annotation: Capture standard output, errors and database manifests. 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. Audit one variant end to end
Trace one allele from VCF through conversion, gene annotation, region overlap, filter match and final multianno columns.
Working checkpoint — region-based annotation: Use the trace as a pipeline acceptance test. 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
ANNOVAR assembles annotations from gene models and databases. It does not replace professional classification, phenotype review or laboratory confirmation.
Working checkpoint — filter-based annotation: Write “matched the dated ClinVar record” instead of “is pathogenic”. 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 three-station classroom activity can sort the same card by membership, location and exact match. Each station answers a different question.
Working checkpoint — protocol operation: Use invented coloured shapes rather than health data. 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
Gene structure, regulatory regions and population variation show that one DNA change can be described at several biological layers.
Working checkpoint — database release: Label which layer each annotation belongs to. 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
Sets, interval overlap and lookup tables make the annotation logic visible. A match is meaningful only with its key and reference universe.
Working checkpoint — independent evidence: Ask students to define the matching rule before counting hits. 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
Data conversion, indexing, protocol arrays and database versioning are core scientific operations. A shifted column can change every interpretation.
Working checkpoint — gene-based annotation: Validate schemas and row counts automatically. 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
Functional annotation joins genetics, databases, statistics and software engineering and can illuminate STEM education and career pathways.
Working checkpoint — region-based annotation: Use current official institutions for real pathway information. 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?
ANNOVAR deliberately separates gene-, region- and filter-based annotation. That design reminds readers that “what gene?”, “what interval?” and “what database match?” are different questions.
Working checkpoint — filter-based annotation: Name the question before reading the answer. 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 the result responsibly
A strong report lists build, input representation, protocol, operation, database version, transcript, allele-specific result and validation status.
Working checkpoint — protocol operation: Keep original evidence fields beside any simplified label. 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 coordinates, audit databases, compare annotation engines, review population and clinical context, then add RNA, protein or functional evidence as the claim requires.
Working checkpoint — database release: Stop at annotation when validation has not been performed. 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 VCF and avinput, conversion command, ANNOVAR release, protocol and operation strings, database manifest, complete multianno output, logs, filters and validation record.
Working checkpoint — independent evidence: Make each interpretation reconstructable from the dated source rows. 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