eduKateSG · Why Science?
Build a cohort network between mutations and RNA events—and demand recurrence, controls and validation before calling causation
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. SAVNet is software for extracting splicing-associated variants from somatic-mutation data together with aberrant splice-junction and intron-retention evidence. It builds cohort-level links between variants and splicing changes, as used in a large exome–transcriptome analysis of human cancers. The result is an association supported by recurrence and modelling across samples; it does not by itself prove that one mutation caused the RNA event or that the event drives disease. Responsible analysis therefore keeps carrier counts, informative non-carriers, gene expression, RNA quality, tumour purity, distance rules and complete tested-edge families visible. A nearby variant, a plausible motif and a recurrent junction make a strong candidate; allele-aware RNA, independent replication and perturbation supply progressively stronger evidence. This layered view is exactly why the method is educational: it shows how DNA and RNA measurements can be connected without collapsing association into causation. It also makes negative evidence useful: absent junctions in genuinely informative controls narrow the story, whereas missing reads in low-expression samples do not. 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: RegTools and variant-aware splice-junction analysis; IRFinder and intron-retention detection; JuncBASE and junction-based splicing events; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official SAVNet repository; Official SAVNet 0.4.0 documentation; Primary Genome Research study; Primary study full text; 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 |
|---|---|---|---|
| Collect | Which samples have somatic variants and RNA events? | Matched mutation, junction, retention and metadata files | A harmonised cohort evidence matrix |
| Link | Which nearby variants and events co-occur? | Genomic proximity, sample membership and event classes | Candidate variant–splicing edges |
| Evaluate | Is the association stronger than expected? | Cohort recurrence, controls and model evidence | Prioritised splicing-associated variants |
| Validate | Does perturbation or independent RNA support causation? | Orthogonal functional evidence | A causal claim only when separately demonstrated |
Inside this guide
1–12 · Foundations and design
- 1. Define a splicing-associated variant
- 2. Use matched DNA and RNA
- 3. Name the biological unit
- 4. Prepare somatic mutations
- 5. Prepare splice junctions
- 6. Prepare intron retention
- 7. Harmonise genome builds
- 8. Define proximity
- 9. Build the sample-by-event pattern
- 10. Understand the network view
- 11. Classify splicing changes
- 12. Use pooled controls
13–24 · Models and estimation
- 13. Evaluate recurrence
- 14. Use permutation logic
- 15. Control multiple testing
- 16. Distinguish cis association
- 17. Consider tumour purity
- 18. Consider subclonality
- 19. Consider expression
- 20. Consider RNA quality
- 21. Inspect splice motifs
- 22. Inspect genome-browser evidence
- 23. Test sample influence
- 24. Test caller sensitivity
25–36 · Diagnostics and interpretation
- 25. Replicate in another cohort
- 26. Perturb the candidate variant
- 27. Connect to cancer biology carefully
- 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. Use bounded result language
- 35. Build a causal ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Define a splicing-associated variant
SAVNet seeks variants whose sample pattern is associated with aberrant junction or intron-retention events. The term SAV is an analysis result, not an automatic statement of mechanism.
Working checkpoint — splicing-associated variant: Write the variant, event and cohort-level association separately. 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. Use matched DNA and RNA
Somatic mutation calls and transcript evidence must come from the same biological samples. A swapped identifier can create a convincing but false co-occurrence.
Working checkpoint — somatic mutation: Reconcile patient, specimen and aliquot identifiers before modelling. 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. Name the biological unit
A tumour specimen is an independent unit only within the sampling design. Multiple aliquots, lanes or libraries from one tumour should not silently inflate recurrence.
Working checkpoint — aberrant junction: Create a specimen map and decide how repeated samples are handled. 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. Prepare somatic mutations
Mutation calls depend on matched normal data, callers, filters and genome build. False somatic calls and missing subclones propagate into the association network.
Working checkpoint — intron retention: Preserve caller versions, filters, allele fractions and quality fields. 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. Prepare splice junctions
Split RNA-seq reads support exon–exon connections. Novel or altered junctions require consistent extraction, strandedness and mapping-quality rules across the cohort.
Working checkpoint — cohort recurrence: Keep junction coordinates, strand and supporting-read counts. 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. Prepare intron retention
SAVNet can also use intron-retention evidence, which depends on intronic coverage, boundary reads and expression context. DNA contamination can imitate retention.
Working checkpoint — permutation evidence: Document retention extraction and contamination checks. 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. Harmonise genome builds
Mutation coordinates, junctions, introns and gene annotation must inhabit the same reference assembly. Lift-over can alter boundaries or fail in complex regions.
Working checkpoint — splicing-associated variant: Reject mixed-build inputs or document validated conversion. 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. Define proximity
Candidate variants are connected to nearby splicing events under configurable spatial rules. A broad window increases coverage and coincidental edges; a narrow window can miss regulatory effects.
Working checkpoint — somatic mutation: Record the window and test a justified alternative. 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. Build the sample-by-event pattern
The model needs to know which samples carry each variant and which display each splicing event. Sparse binary or weighted patterns make recurrence visible.
Working checkpoint — aberrant junction: Audit a few rows against raw files before fitting. 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. Understand the network view
Variants and splicing events form nodes linked when their genomic and sample evidence make an association plausible. Network structure helps pool evidence across related events and samples.
Working checkpoint — intron retention: Interpret edges as statistical relationships, not physical binding. 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. Classify splicing changes
The documentation distinguishes categories of splicing changes and splicing-associated variants. Classification depends on junction structures and annotation.
Working checkpoint — cohort recurrence: Keep the raw coordinates beside every category label. 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. Use pooled controls
Pooled normal or reference information can help identify junctions that are recurrent technical or physiological features rather than tumour-specific aberrations.
Working checkpoint — permutation evidence: Match processing and tissue context when constructing controls. 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. Evaluate recurrence
A variant–event association seen across multiple independent tumours can be more persuasive than a singleton. Yet recurrent sequencing artefacts can also repeat.
Working checkpoint — splicing-associated variant: Show counts, cohort denominator and technical covariates. 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. Use permutation logic
Rearranging sample labels or comparable null procedures can ask whether observed co-occurrence exceeds chance under an explicit null. Exchangeability assumptions must be plausible.
Working checkpoint — somatic mutation: Describe exactly what was permuted and what structure was preserved. 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. Control multiple testing
Large cohorts contain many variants, junctions and possible edges. Correction and filtering families must match the reported claims.
Working checkpoint — aberrant junction: Retain the complete tested network and adjusted evidence. 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 cis association
A nearby variant associated with a splice event is a cis-acting candidate. Proximity and co-occurrence do not prove that the allele altered spliceosome recognition.
Working checkpoint — intron retention: Use allele-aware or perturbational evidence when claiming cis action. 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. Consider tumour purity
Low tumour purity can weaken both somatic allele fraction and tumour-specific RNA signal. Variable purity may create heterogeneous detectability across samples.
Working checkpoint — cohort recurrence: Report purity estimates or sensitivity analyses where relevant. 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. Consider subclonality
A mutation in a small tumour subclone may produce a weak RNA signature even if biologically active. Conversely, abundant aberrant RNA may come from another process.
Working checkpoint — permutation evidence: Compare variant allele fraction with event support cautiously. 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. Consider expression
An event cannot be detected reliably when the host gene has little informative RNA. Negative evidence must be interpreted against expression and depth.
Working checkpoint — splicing-associated variant: Display gene expression and local usable reads. 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. Consider RNA quality
Degradation, library preparation and batch can alter junction and intron evidence. If these factors correlate with tumour subtype, association can be confounded.
Working checkpoint — somatic mutation: Use metadata plots and matched processing whenever possible. 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. Inspect splice motifs
A variant near canonical donor or acceptor bases may provide a plausible sequence explanation. Deep exonic or intronic candidates may require enhancer, silencer or cryptic-site hypotheses.
Working checkpoint — aberrant junction: Keep motif plausibility distinct from cohort association strength. 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. Inspect genome-browser evidence
Read-level views can reveal alignment artefacts, nearby indels and whether all supporting junctions fit one coherent event.
Working checkpoint — intron retention: Use identical scales and display multiple carriers plus controls. 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. Test sample influence
A network edge can depend on one high-depth or unusual tumour. Leave-one-sample-out analysis makes this fragility visible.
Working checkpoint — cohort recurrence: Flag edges whose evidence collapses with one specimen removed. 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. Test caller sensitivity
Different somatic and junction callers can change candidate inputs. Agreement under defensible pipelines increases confidence but does not prove causality.
Working checkpoint — permutation evidence: Predefine comparisons and report disagreements. 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. Replicate in another cohort
Independent tumour cohorts test whether the same variant–event relationship generalises across sampling and processing.
Working checkpoint — splicing-associated variant: Repeat the primary definition before exploring new thresholds. 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. Perturb the candidate variant
Genome editing or minigene experiments can test whether changing the allele alters splicing under controlled conditions. Model systems may still differ from tumour tissue.
Working checkpoint — somatic mutation: Design wild-type and alternative constructs with matched context. 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. Connect to cancer biology carefully
Some splicing-associated variants may affect tumour genes or pathways, but association alone does not establish driver status, treatment sensitivity or prognosis.
Working checkpoint — aberrant junction: Require functional and clinical evidence for stronger claims. 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 fair comparison asks whether samples with one feature also show another more often than expected while controls and conditions remain visible.
Working checkpoint — intron retention: Use harmless classroom cards to model association, not cancer diagnosis. 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
Mutation, transcription and RNA processing show how DNA and RNA measurements can be connected, but a measured relationship still needs causal testing.
Working checkpoint — cohort recurrence: Label each step observed, associated or experimentally tested. 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
Contingency patterns, networks, probabilities and multiple comparisons explain why a repeated association can still occur by chance.
Working checkpoint — permutation evidence: Use small invented matrices to calculate simple co-occurrence. 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
Coordinate joins, sparse matrices and workflow versions are experimental apparatus. One sample-label or strand error can manufacture an edge.
Working checkpoint — splicing-associated variant: Make identity and build checks stop the workflow. 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
Cancer genomics joins molecular biology, statistics, bioinformatics and software engineering and can illuminate STEM education and career pathways.
Working checkpoint — somatic mutation: Use official institutions for current programme 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?
SAVNet can combine two RNA aberration types—splice junction changes and intron retention—with somatic variants across a cohort. This widens the event catalogue while increasing the need for clear denominators.
Working checkpoint — aberrant junction: Ask how many samples were informative for each edge. 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. Use bounded result language
Write that SAVNet identified a splicing-associated variant under stated cohort and filtering rules. Avoid “caused” or “driver” until independent evidence supports those verbs.
Working checkpoint — intron retention: Put validation status in the main results table. 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 causal ladder
Move from co-occurrence to sequence plausibility, allele-aware RNA, replication and perturbation. Each rung answers a stronger question.
Working checkpoint — cohort recurrence: Stop the conclusion at the highest completed rung. 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 matched mutation, junction and retention inputs, sample map, reference files, SAVNet version and configuration, complete edges, null analyses, plots and validation status.
Working checkpoint — permutation evidence: Make every SAV traceable to its carrier and informative controls for independent review. 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