eduKateSG · Why Science?
Read exons and splice junctions together when an annotation may not tell the whole story
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. JunctionSeq is an RNA-seq method that extends DEXSeq-style differential-usage modelling to both exonic regions and splice junctions. It can test annotated features and junctions discovered in aligned reads, so an incomplete transcript catalogue need not hide every alternative pattern. Its output supports a local usage claim under a specified design; it does not by itself reconstruct a full isoform or prove a mechanism. 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: DEXSeq and differential exon usage; SUPPA2 and uncertainty-aware differential splicing; LeafCutter and annotation-free intron excision; MAJIQ and local splicing variations; rMATS and replicate-aware differential splicing; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: JunctionSeq primary study in Nucleic Acids Research; JunctionSeq primary study record in PubMed; Official JunctionSeq project documentation; 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 |
|---|---|---|---|
| Count | Which exonic regions and junctions receive read support? | Splice-aware alignments, annotation and QoRTs-derived counts | A versioned feature-count table |
| Model | Does a feature change relative to its gene? | Independent biological replicates and a declared design | Differential-usage evidence |
| Inspect | Which annotated or novel feature carries the pattern? | Gene profiles, counts and coordinates | A local alternative-usage hypothesis |
| Validate | Does another measurement support the structure? | Browser review, targeted RNA assays or long reads | A bounded, reproducible conclusion |
Inside this guide
1–12 · Foundations and design
- 1. Ask about usage, not only expression
- 2. Begin with a splice-aware alignment
- 3. Freeze the transcript annotation
- 4. Use QoRTs for count preparation
- 5. Understand exonic regions
- 6. Understand junction features
- 7. Distinguish known from novel
- 8. Name the independent replicate
- 9. Build the design before looking at hits
- 10. Model the gene–feature interaction
- 11. Estimate dispersion across replicates
- 12. Filter weak evidence transparently
13–24 · Models and estimation
- 13. Keep gene expression separate
- 14. Correct for multiple testing
- 15. Read effect size with adjusted evidence
- 16. Use gene-level summaries carefully
- 17. Read the gene profile as a map
- 18. Inspect the dashed novel junctions
- 19. Check coordinate conventions
- 20. Check strandedness
- 21. Check multimapping and repeats
- 22. Check one influential sample
- 23. Check annotation sensitivity
- 24. Check alignment sensitivity
25–36 · Diagnostics and interpretation
- 25. Group overlapping feature signals
- 26. Compare with DEXSeq fairly
- 27. Compare with LeafCutter fairly
- 28. Compare with full-transcript methods
- 29. Connect to fair-test habits
- 30. Connect to graphs and maps
- 31. Connect to computing
- 32. Connect to education pathways
- 33. Use bounded scientific language
- 34. Build a validation ladder
- 35. Did you know?
- 36. Audit one gene end to end
37–37 · Validation, learning and pathways
Section 1 of 37
1. Ask about usage, not only expression
JunctionSeq compares each exon or splice junction with the expression of its parent gene. A feature can therefore be differentially used even when its own count looks stable, because the gene total and competing features define the denominator.
Working checkpoint — exonic region: Write the feature-versus-gene question before running a 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 2 of 37
2. Begin with a splice-aware alignment
The workflow relies on reads aligned across exon boundaries as well as reads assigned to exonic regions. Genome build, aligner settings and handling of multimapping reads all influence the feature evidence.
Working checkpoint — splice junction: Archive the reference genome and complete alignment command. 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 37
3. Freeze the transcript annotation
An annotation supplies known exons, transcripts and junctions, while observed alignments can add novel junctions. Changing the annotation can change bins, gene membership and which features are labelled novel.
Working checkpoint — differential usage: Record release, source and chromosome naming conventions. 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 37
4. Use QoRTs for count preparation
The primary JunctionSeq workflow uses QoRTs to generate exon and splice-junction counts and related quality-control information. Count preparation is part of the analysis, not a disposable prelude.
Working checkpoint — novel junction: Save the QoRTs version, parameters and logs. 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 37
5. Understand exonic regions
Overlapping transcript exons are divided into countable exonic regions so feature usage can be compared within a gene. These bins are computational summaries and need not equal biological exons named in a textbook.
Working checkpoint — alternative isoform regulation: Show coordinates and transcript overlaps for a representative region. 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 37
6. Understand junction features
A junction feature represents reads that connect two genomic splice boundaries. It is direct evidence for a local connection, but it does not reveal every upstream and downstream exon on the same RNA molecule.
Working checkpoint — gene profile: Avoid treating one junction as a complete isoform. 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 37
7. Distinguish known from novel
Known junctions appear in the supplied annotation; novel junctions are supported by alignments but absent from that catalogue. Novel means unannotated in that reference, not newly evolved or automatically functional.
Working checkpoint — exonic region: State the annotation release whenever using the word novel. 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 37
8. Name the independent replicate
Biological replicates are independently sampled organisms, people, cultures or specimens. Additional lanes and millions of reads can improve measurement without increasing the number of independent experimental units.
Working checkpoint — splice junction: Put the experimental unit in the sample table. 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 37
9. Build the design before looking at hits
Condition, batch, pairing and interactions belong in a design that reflects how samples were collected. A formula cannot separate condition from a perfectly confounded batch.
Working checkpoint — differential usage: Draw the design matrix and check its rank. 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 37
10. Model the gene–feature interaction
Like DEXSeq, JunctionSeq asks whether the condition effect differs for a feature relative to the gene-wide pattern. The interaction term is the statistical expression of differential usage.
Working checkpoint — novel junction: Explain the model term in plain language beside the formula. 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 37
11. Estimate dispersion across replicates
Count variability exceeds simple sampling noise in biological data. Dispersion estimation uses replicate information to keep a dramatic but inconsistent feature from looking more precise than it is.
Working checkpoint — alternative isoform regulation: Plot sample-level counts and inspect dispersion diagnostics. 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 37
12. Filter weak evidence transparently
Features with extremely low counts cannot support stable comparisons. Filtering can improve inference, but it also changes the tested family and must be reported.
Working checkpoint — gene profile: Save before-and-after feature counts and every threshold. 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 37
13. Keep gene expression separate
A strongly changing gene can make non-changing features look relatively different, and a stable gene can contain a clear usage shift. Differential expression and differential usage are complementary questions.
Working checkpoint — exonic region: Show gene totals next to feature estimates. 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 37
14. Correct for multiple testing
Thousands of exonic regions and junctions may be tested. Adjusted values belong to the declared family under the chosen filters and cannot be interpreted as universal error probabilities.
Working checkpoint — splice junction: Preserve the complete result table, including non-significant rows. 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 37
15. Read effect size with adjusted evidence
A small adjusted value does not say whether the usage shift is biologically large. Conversely, a large estimated shift from sparse counts can remain unreliable.
Working checkpoint — differential usage: Report counts, effect direction and adjusted evidence together. 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 37
16. Use gene-level summaries carefully
JunctionSeq can aggregate feature evidence to help prioritise genes. A gene-level signal says at least part of the local profile differs; it does not identify the causal transcript or regulator.
Working checkpoint — novel junction: Open the gene profile before naming the biological event. 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 37
17. Read the gene profile as a map
The plotting system brings exonic regions, junctions, transcripts and normalized expression into one coordinated view. Its value is interpretive: it helps readers see where the statistical signal sits.
Working checkpoint — alternative isoform regulation: Export the figure and the underlying numeric table. 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 37
18. Inspect the dashed novel junctions
In JunctionSeq plots, annotated and novel connections can be visually distinguished. A dashed connection deserves mapping review because repetitive sequence, paralogues or alignment artefacts can imitate novelty.
Working checkpoint — gene profile: Inspect split-read placement in a genome browser. 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 37
19. Check coordinate conventions
Different tools may use zero-based or one-based endpoints and different rules for intron boundaries. A one-base mismatch can make a known junction appear novel.
Working checkpoint — exonic region: Hand-check several canonical junctions through the pipeline. 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 37
20. Check strandedness
A wrong library-strand setting can assign evidence to the wrong gene or reverse donor and acceptor interpretation. Strand information must agree among protocol, alignments, counts and annotation.
Working checkpoint — splice junction: Verify strandedness empirically and document the result. 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 37
21. Check multimapping and repeats
Reads that map equally well to several loci can inflate or blur local evidence. The aligner and counting rules determine whether such reads are excluded, down-weighted or assigned.
Working checkpoint — differential usage: Flag decisive features in repetitive sequence. 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 37
22. Check one influential sample
A significant interaction can be carried by one specimen. Leave-one-replicate-out fits and per-sample plots expose fragility without pretending to create new data.
Working checkpoint — novel junction: Lower confidence when direction changes repeatedly. 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 37
23. Check annotation sensitivity
A justified alternative annotation may convert a novel junction into an annotated one or reorganise exonic regions. Robust biological patterns should be discussed across those changes.
Working checkpoint — alternative isoform regulation: Repeat headline loci with one planned catalogue sensitivity. 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 37
24. Check alignment sensitivity
Splice-aware aligners differ in seed, overhang and novel-junction rules. Agreement under a defensible alternative is reassuring; disagreement identifies a measurement boundary.
Working checkpoint — gene profile: Keep genome and annotation fixed during the comparison. 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 37
25. Group overlapping feature signals
Several significant rows can describe one local splice change because exons and junctions share structure. Counting every row as a separate discovery exaggerates the biology.
Working checkpoint — exonic region: Group features by locus and shared splice paths. 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 37
26. Compare with DEXSeq fairly
DEXSeq focuses on exonic-region usage, whereas JunctionSeq adds junction features and improved profiles. Their hypotheses overlap but are not identical when unannotated connections matter.
Working checkpoint — splice junction: Compare matched regions and designs, not raw hit counts. 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 37
27. Compare with LeafCutter fairly
LeafCutter clusters excised introns from junction reads without requiring transcript annotation for discovery. JunctionSeq combines exon and junction usage relative to genes.
Working checkpoint — differential usage: Translate each result into its native unit before comparison. 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 37
28. Compare with full-transcript methods
Transcript-level tools try to estimate complete isoform mixtures, while JunctionSeq makes local feature claims. A local result can be strong even when full transcript phasing remains unresolved.
Working checkpoint — novel junction: Use long reads when the headline depends on complete isoforms. 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 37
29. Connect to fair-test habits
The advanced model still asks familiar questions: what changed, what stayed comparable, how many independent repeats were used and which alternative explanation remains?
Working checkpoint — alternative isoform regulation: Practise these questions in Primary Science and Secondary Science. 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 37
30. Connect to graphs and maps
Junction arcs and exon boxes turn a molecule into a navigable structure. The representation connects biology with network thinking and visual literacy.
Working checkpoint — gene profile: Ask learners to trace alternative routes without assuming every route was observed end to end. 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 37
31. Connect to computing
Identifiers, genomic intervals, joins and reproducible scripts are experimental apparatus. A correct model applied to shifted coordinates can produce confidently wrong output.
Working checkpoint — exonic region: Add assertions for genome build, chromosome names and sample identity. 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 37
32. Connect to education pathways
The work joins molecular biology, statistics and computing and can illuminate routes into life science, biotechnology and bioinformatics. It does not guarantee admission or employment.
Working checkpoint — splice junction: Use current official programme pages 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 37
33. Use bounded scientific language
A result supports differential usage of a stated feature under a model and annotation. It does not alone prove a protein change, clinical meaning or causal mechanism.
Working checkpoint — differential usage: Match each verb to the measurement actually made. 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 37
34. Build a validation ladder
Browser inspection challenges mapping; targeted RNA assays challenge the local structure; long reads challenge phasing; protein and functional assays challenge consequences.
Working checkpoint — novel junction: Validate the weakest link in the proposed claim first. 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 37
35. Did you know?
JunctionSeq was designed to retain sensitivity to alternative isoform regulation even when the annotation is incomplete, because observed splice junctions can enter the test without a separate transcript-assembly step.
Working checkpoint — alternative isoform regulation: Treat discovery of a junction and interpretation of an isoform as different stages. 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 37
36. Audit one gene end to end
Trace one gene from annotation and BAM files through QoRTs counts, feature definitions, design, fitted model, adjusted values and gene profile.
Working checkpoint — gene profile: An end-to-end audit is the fastest way to expose hidden assumptions. 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 37 of 37
37. Finish with a reproducible bundle
Deliver annotation, genome build, alignments or accessions, QoRTs and JunctionSeq versions, metadata, design, filters, complete results, plots and sensitivity checks.
Working checkpoint — exonic region: A reader should be able to reproduce both the shortlist and its exclusions. 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.
Contents · Previous section · Continue to the Science Learning Hub