eduKateSG · Why Science?
Measure an isoform mixture, model its uncertainty and ask a sharper splicing question
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. SUPPA2 is a software framework for studying alternative splicing from transcript-abundance estimates. It can define local splicing events from an annotation, calculate Percent Spliced In (PSI) values and test whether PSI changes across conditions while relating the observed change to variability among biological replicates. It supports transcript-level as well as local-event analysis, but its answers inherit uncertainty from annotation, quantification and experimental design. 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; EventPointer and interpretable splicing events; DRIMSeq and transcript usage; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official SUPPA repository and current v2.4 release information; SUPPA2 primary study in Genome Biology; SUPPA2 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 |
|---|---|---|---|
| Define | Which local event or transcript mixture is being measured? | GTF annotation and event-generation rules | A versioned event catalogue |
| Quantify | What share supports inclusion? | Transcript abundance, commonly TPM | PSI per sample with a visible denominator |
| Compare | Is the observed change unusual for its expression range? | Condition contrast and biological replicates | Delta-PSI plus statistical evidence |
| Validate | Does the splice pattern survive another measurement? | Junction views, sensitivity checks and targeted evidence | A supported lead, not an automatic mechanism |
Inside this guide
1–12 · Foundations and design
- 1. Begin with the biological choice
- 2. Read PSI as a proportion
- 3. Separate PSI from expression
- 4. Generate events from a frozen GTF
- 5. Recognise the supported event families
- 6. Treat alternative ends carefully
- 7. Prepare transcript abundance consistently
- 8. Understand abundance allocation
- 9. Keep identifiers aligned
- 10. Identify the independent replicate
- 11. Calculate PSI sample by sample
- 12. Handle low-expression events
13–24 · Models and estimation
- 13. Define delta-PSI direction
- 14. Use replicate variability as evidence
- 15. Choose the differential method explicitly
- 16. Do not confuse a p-value with effect size
- 17. Correct the tested family
- 18. Inspect event dependence
- 19. Plot every biological replicate
- 20. Check junction support
- 21. Challenge annotation dependence
- 22. Challenge quantifier dependence
- 23. Challenge sample influence
- 24. Separate local events from full isoforms
25–36 · Diagnostics and interpretation
- 25. Compare with count-based methods fairly
- 26. Connect to fair-testing habits
- 27. Connect to fractions and denominators
- 28. Connect to computing
- 29. Connect to education pathways
- 30. Use bounded language
- 31. Build a validation ladder
- 32. Report negative and ambiguous results
- 33. Audit one event end to end
- 34. Translate for a non-specialist
- 35. Did you know?
- 36. Preserve the software environment
37–37 · Validation, learning and pathways
Section 1 of 37
1. Begin with the biological choice
Alternative splicing lets one pre-mRNA contribute to different mature transcript structures. SUPPA2 turns that broad idea into defined transcript or local-event questions, so the first responsibility is to name the exact choice being compared.
Working checkpoint — percent spliced in: Draw the included and excluded structures before computing a PSI value. 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. Read PSI as a proportion
Percent Spliced In is a proportion assigned to an event or transcript feature within a defined mixture. A PSI near one means the inclusion form dominates that quantified mixture; it does not mean every cell contains it.
Working checkpoint — transcript abundance: State the numerator, denominator and sample whenever presenting PSI. 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. Separate PSI from expression
A transcript can have a stable relative share while the gene changes abundance, or its share can change while the gene total stays similar. Splicing and expression are related but non-identical questions.
Working checkpoint — alternative-splicing event: Show abundance beside PSI so a changing denominator is not hidden. 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. Generate events from a frozen GTF
SUPPA2 can derive local alternative-splicing events from a transcript annotation in GTF format. The event catalogue therefore depends on exon coordinates, transcript membership and annotation release.
Working checkpoint — between-replicate variability: Archive the exact GTF and event-generation command with the results. 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. Recognise the supported event families
The documented workflow covers exon skipping, alternative 5-prime and 3-prime splice sites, mutually exclusive exons, retained introns and alternative first or last exons. Each event type has its own structural interpretation.
Working checkpoint — differential splicing: Report event type and coordinates instead of relying on an opaque identifier. 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. Treat alternative ends carefully
Alternative first and last exons can reflect promoter or polyadenylation choices as well as splicing-related structure. A computational event label does not establish the regulatory mechanism.
Working checkpoint — annotation release: Use capped or long-read evidence when a claim depends on transcript ends. 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. Prepare transcript abundance consistently
SUPPA2 commonly calculates event PSI from transcript abundance estimates such as TPM. All samples should come from the same reference, quantifier settings and transcript catalogue.
Working checkpoint — percent spliced in: Reject a comparison assembled from incompatible transcriptomes or abundance units. 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. Understand abundance allocation
Short reads shared by several isoforms create ambiguity. A quantifier distributes evidence under a statistical model, and SUPPA2 inherits those estimates rather than observing each molecule directly.
Working checkpoint — transcript abundance: Inspect equivalence or uncertainty information for decisive 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 9 of 37
9. Keep identifiers aligned
Transcript identifiers in abundance files must match those in the GTF used to define events. Version suffixes and retired identifiers can silently remove components from a PSI calculation.
Working checkpoint — alternative-splicing event: Count matched, unmatched and duplicated identifiers before analysis. 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. Identify the independent replicate
Independent people, organisms, cultures or specimens support biological inference. Technical lanes and repeated software runs can improve measurement but do not create new biological replicates.
Working checkpoint — between-replicate variability: Write the experimental unit into the sample sheet. 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. Calculate PSI sample by sample
The psiPerEvent workflow combines abundances of transcripts supporting alternative forms. A valid-looking number still depends on the event definition and availability of its component transcripts.
Working checkpoint — differential splicing: Audit one event by tracing every contributing transcript and abundance. 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. Handle low-expression events
PSI becomes unstable when the underlying expression is extremely small. SUPPA2 links differential-splicing assessment to expression because low-abundance measurements often show greater apparent variability.
Working checkpoint — annotation release: Apply and report a defensible expression filter before interpretation. 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. Define delta-PSI direction
Delta-PSI is the difference between condition summaries in a specified order. Reversing the contrast reverses the sign without changing the biology.
Working checkpoint — percent spliced in: Print the subtraction order in every figure legend. 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. Use replicate variability as evidence
SUPPA2 compares an observed between-condition PSI difference with PSI differences seen among biological replicates at similar expression. The aim is to calibrate change against empirical variability rather than use magnitude alone.
Working checkpoint — transcript abundance: Preserve replicate pairs and the expression binning or pooling settings. 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. Choose the differential method explicitly
The software provides empirical testing and, for larger replicate sets, a Wilcoxon option described in its documentation. These approaches are not interchangeable defaults.
Working checkpoint — alternative-splicing event: Record the selected method, version and every threshold. 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. Do not confuse a p-value with effect size
A small probability value can accompany a modest delta-PSI, while a large shift can remain uncertain in a small experiment. Statistical evidence and practical magnitude answer different questions.
Working checkpoint — between-replicate variability: Report delta-PSI, p-value, adjusted value and sample count together. 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. Correct the tested family
Thousands of events can be evaluated. Multiple-testing correction addresses the declared family after filtering; changing that family changes the interpretation of adjusted values.
Working checkpoint — differential splicing: Save the full event table and the exact pre-test filtering log. 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 event dependence
Events from the same gene can overlap and share transcripts, so their tests are not biologically independent stories. Multiple significant rows may describe one structural rearrangement.
Working checkpoint — annotation release: Group related events by locus before counting biological findings. 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. Plot every biological replicate
Condition means can hide outliers, bimodality or a batch pattern. Sample-level PSI and expression plots reveal whether a result is shared or carried by one specimen.
Working checkpoint — percent spliced in: Place individual points over summaries and label influential samples. 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 junction support
Transcript-derived PSI is valuable, but a genome-browser or junction-count view can reveal whether the relevant splice boundaries have direct read support.
Working checkpoint — transcript abundance: Inspect inclusion and exclusion junctions for headline events. 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. Challenge annotation dependence
A different justified annotation may split, merge or omit transcripts and thereby change event definitions and PSI. Annotation sensitivity is part of robustness, not a nuisance to conceal.
Working checkpoint — alternative-splicing event: Repeat key events with one planned alternative catalogue. 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. Challenge quantifier dependence
Different transcript quantifiers can allocate ambiguous reads differently. Agreement is reassuring; disagreement identifies an uncertainty source that deserves explanation.
Working checkpoint — between-replicate variability: Use identical references and report the sensitivity 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 23 of 37
23. Challenge sample influence
A leave-one-replicate-out analysis can show whether delta-PSI direction and significance depend on one biological unit. It is a fragility check, not a replacement for replication.
Working checkpoint — differential splicing: Flag events that repeatedly change conclusion when one sample is 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 37
24. Separate local events from full isoforms
A local exon-skipping event does not specify the complete transcript carrying it. Conversely, transcript-level PSI combines many structural choices into a full-isoform proportion.
Working checkpoint — annotation release: Match the reporting unit to the biological question. 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. Compare with count-based methods fairly
LeafCutter, MAJIQ, rMATS and exon-usage methods use different units, inputs and uncertainty models. Their output rows should not be compared as if they were identical tests.
Working checkpoint — percent spliced in: Align samples, genome build and question before comparing calls. 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. Connect to fair-testing habits
The advanced workflow still depends on familiar science habits: change one comparison, use independent repeats, measure consistently and consider alternative explanations.
Working checkpoint — transcript abundance: Ask students to identify the variable, control and replicate before the algorithm. 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. Connect to fractions and denominators
PSI is a ratio. Like a class percentage or recipe share, it can move because one component changes, another changes or the total composition shifts.
Working checkpoint — alternative-splicing event: Have learners reconstruct a PSI value from coloured counters. 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. Connect to computing
GTF parsing, coordinate systems, vectorised abundance tables and reproducible commands make the analysis possible. Computing literacy protects the biological identities carried by each row.
Working checkpoint — between-replicate variability: Add assertions for sample names, coordinate conventions and transcript matches. 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 education pathways
Alternative-splicing analysis sits at the intersection of biology, statistics and computing and can illuminate routes into life science, biotechnology and bioinformatics.
Working checkpoint — differential splicing: Check current official course information rather than treating one tool as an entry requirement. 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. Use bounded language
A SUPPA2 result supports a quantified difference under a defined annotation and model. It does not alone prove protein change, phenotype, diagnosis or treatment benefit.
Working checkpoint — annotation release: Use “supports differential splicing” rather than “causes disease” unless separate evidence warrants it. 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. Build a validation ladder
Junction inspection tests read support; targeted RNA assays test the splice form; long reads test transcript structure; protein and functional assays test downstream consequences.
Working checkpoint — percent spliced in: Choose the next rung according to the claim, not convenience. 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. Report negative and ambiguous results
Events filtered for low expression or unstable PSI remain part of the evidence record. Absence of significance is not proof that splicing is identical.
Working checkpoint — transcript abundance: Distinguish no detected difference from evidence of equivalence. 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. Audit one event end to end
Trace one event from GTF coordinates through supporting transcripts, sample abundances, PSI, delta-PSI, replicate differences, adjustment and plots.
Working checkpoint — alternative-splicing event: An end-to-end audit makes silent assumptions visible. 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. Translate for a non-specialist
A clear explanation says which splice choice changed, by how much, across how many independent samples and with what uncertainty.
Working checkpoint — between-replicate variability: Avoid compressing the whole chain into “the gene was switched”. 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?
SUPPA2 uses transcript abundance to calculate many event PSI values quickly, so one quantified transcriptome can support both local-event and transcript-level views. Speed expands the questions; it does not remove quantification uncertainty.
Working checkpoint — differential splicing: Treat computational efficiency and evidential strength as separate qualities. 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. Preserve the software environment
The official repository identifies SUPPA v2.4 as the current release checked for this article. Dependencies, commands and output formats still belong in the methods archive.
Working checkpoint — annotation release: Save the version string, environment and command history. 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 the GTF, event catalogue, transcript abundances, sample metadata, contrast, filters, commands, complete results, plots, sensitivity checks and narrowest conclusion.
Working checkpoint — percent spliced in: A reader should be able to reproduce both the shortlist and the reasons a result was excluded. 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