eduKateSG · Why Science?
Bring splicing, expression and sample context into one workspace—then keep exploration separate from confirmation
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. psichomics is an R/Bioconductor package with command-line and Shiny interfaces for quantifying, analysing and visualising alternative splicing together with gene expression and sample-associated information. Official documentation describes support for TCGA, GTEx, selected SRA resources, VAST-TOOLS outputs and user data, with differential, dimensionality-reduction, correlation and survival analyses. Its strength is integrative exploration; responsible use keeps dataset provenance, grouping rules, missingness and multiple testing visible and treats exploratory associations as hypotheses. 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: vast-tools and cross-sample splicing profiles; SpliceWiz and interactive alternative-splicing analysis; SGSeq and genome-guided event prediction; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official Bioconductor psichomics release page; Official psichomics documentation; psichomics primary article in Nucleic Acids Research; psichomics 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 |
|---|---|---|---|
| Load | Which samples, junctions and attributes are available? | Curated or user-provided datasets | A versioned integrated workspace |
| Quantify | What fraction supports each event? | Inclusion and exclusion splice-junction counts | PSI estimates with missingness |
| Explore | Which patterns align with groups or outcomes? | Differential, dimensional, correlation and survival tools | Transparent associations and prioritised events |
| Confirm | Which findings survive an independent test? | Predeclared models and orthogonal evidence | A bounded conclusion beyond exploration |
Inside this guide
1–12 · Foundations and design
- 1. Begin with an integration question
- 2. Know the supported data routes
- 3. Keep public data provenance
- 4. Prepare user data carefully
- 5. Understand event annotation
- 6. Quantify skipped-exon events
- 7. Quantify other event classes
- 8. Respect unsupported evidence types
- 9. Preserve missing PSI values
- 10. Load gene expression separately
- 11. Group samples from real attributes
- 12. Avoid post-hoc group invention
13–24 · Models and estimation
- 13. Keep subjects and samples distinct
- 14. Run differential splicing transparently
- 15. Interpret median shifts
- 16. Interpret variance shifts
- 17. Use dimensionality reduction as a map
- 18. Handle imputation visibly
- 19. Explore expression–splicing correlation
- 20. Use survival analysis cautiously
- 21. Distinguish prediction from prognosis
- 22. Inspect genomic mapping
- 23. Inspect functional annotations
- 24. Control multiple exploration families
25–36 · Diagnostics and interpretation
- 25. Audit one event end to end
- 26. Test group sensitivity
- 27. Test sample influence
- 28. Compare with vast-tools fairly
- 29. Compare with SpliceWiz fairly
- 30. Connect to school statistics
- 31. Connect to biology
- 32. Connect to computing
- 33. Connect to pathways
- 34. Use bounded language
- 35. Did you know?
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Begin with an integration question
psichomics can place alternative splicing, gene expression and sample annotations in one analysis environment. The scientific question should say which molecular layer, sample group and outcome matter before the convenience of integration expands the search.
Working checkpoint — percent spliced in: Write the primary analysis and exploratory analyses 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. Know the supported data routes
Official documentation describes importing processed TCGA and GTEx resources, selected SRA projects, VAST-TOOLS results and user-provided data. Each route carries different preprocessing and consent or access contexts.
Working checkpoint — splice-junction count: Record dataset release, download date and import route. 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. Keep public data provenance
A familiar cancer or tissue label does not uniquely identify a dataset version. Sample exclusions, metadata revisions and preprocessing updates can change the analysis.
Working checkpoint — sample attribute: Archive accession lists and all inclusion rules. 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 user data carefully
User-provided junction counts, splicing annotation, expression tables and sample attributes must share consistent identifiers. A valid-looking interface can still join the wrong rows.
Working checkpoint — differential splicing: Test dimensions, uniqueness and sample order before loading. 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 event annotation
Splicing quantification requires a catalogue that defines alternative paths and the junctions supporting them. Custom annotation broadens scope but also changes the event universe and denominators.
Working checkpoint — dimensionality reduction: Version the event annotation with the analysis. 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. Quantify skipped-exon events
The official documentation explains PSI from junction reads supporting inclusion and exclusion. For a skipped exon, inclusion paths and the junction that bypasses the exon form the relevant competition.
Working checkpoint — survival association: Display every junction used in the calculation. 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. Quantify other event classes
psichomics supports several junction-defined event types such as mutually exclusive exons and alternative splice sites. The required junction pattern and denominator differ across classes.
Working checkpoint — percent spliced in: Teach the path diagram before comparing PSI values. 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. Respect unsupported evidence types
The primary article notes that junction-based quantification does not directly cover every event type in the same way; intron retention needs intron–exon evidence and alternative untranslated regions may need exon-body evidence.
Working checkpoint — splice-junction count: Do not force all RNA processing into one junction formula. 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. Preserve missing PSI values
A sample may lack enough compatible reads for a stable event estimate. Missingness can correlate with tissue, library depth or expression and therefore deserves analysis rather than zero filling.
Working checkpoint — sample attribute: Report measurement availability by group. 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. Load gene expression separately
Gene expression can be normalised, filtered and transformed in the application or supplied after external processing. The selected route affects scale and comparability.
Working checkpoint — differential splicing: Document units and every transformation. 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. Group samples from real attributes
Clinical and molecular attributes can define groups for exploration. Categories such as stage, tissue or exposure must be checked for missing values, coding changes and small cells.
Working checkpoint — dimensionality reduction: Publish the group-building rule and 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 12 of 36
12. Avoid post-hoc group invention
An interactive tool makes it easy to merge, intersect and redefine groups. Exploring is legitimate, but a group created after seeing molecular separation cannot be presented as a predeclared confirmatory test.
Working checkpoint — survival association: Label exploratory groups visibly. 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. Keep subjects and samples distinct
One person may contribute more than one sample, and repeated measurements are not independent subjects. Survival and differential analyses can be distorted if repeated samples are treated as separate people.
Working checkpoint — percent spliced in: Create a subject-to-sample map. 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. Run differential splicing transparently
The package offers median- and variance-oriented statistical comparisons for splicing and expression. The chosen test, assumptions and multiple-testing family should be named explicitly.
Working checkpoint — splice-junction count: Export test settings with complete 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 15 of 36
15. Interpret median shifts
A difference in central PSI can indicate a systematic usage change, yet similar medians can hide distinct subgroups and different medians can be driven by missingness or one cluster.
Working checkpoint — sample attribute: Plot distributions and individual samples. 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. Interpret variance shifts
Conditions may differ in heterogeneity even when average usage is similar. A variance result describes dispersion of measured PSI, not automatically biological instability.
Working checkpoint — differential splicing: Check depth, batch and missingness before a heterogeneity story. 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 dimensionality reduction as a map
PCA and ICA summarise multivariable patterns. Axes are mathematical combinations that can reveal structure, but separation does not prove that the chosen phenotype caused it.
Working checkpoint — dimensionality reduction: Inspect loadings and possible technical covariates. 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. Handle imputation visibly
The primary article describes filtering variables with too much missing data and imputing remaining values for analyses that cannot accept missingness. Imputation can affect distances and components.
Working checkpoint — survival association: Report threshold, method and sensitivity without imputation. 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. Explore expression–splicing correlation
A correlation between an RNA-binding protein’s expression and an event’s PSI can prioritise regulation hypotheses. Shared cell type, batch or condition can produce the same association.
Working checkpoint — percent spliced in: Plot the variables and adjust only for defensible covariates. 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. Use survival analysis cautiously
Kaplan–Meier curves and Cox models can associate splicing or expression with observed survival. Cut points, censoring, clinical covariates and multiple testing can strongly affect a headline.
Working checkpoint — splice-junction count: Predeclare predictors where possible and validate independently. 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. Distinguish prediction from prognosis
A survival association in a retrospective dataset is not automatically a clinical predictor and does not show that changing the event changes outcome. Clinical utility needs calibration and external validation.
Working checkpoint — sample attribute: Use association language unless prediction was truly evaluated. 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 genomic mapping
Interactive genomic views connect event identifiers to exon and junction structures. Mapping helps interpretation but depends on genome and transcript annotation releases.
Working checkpoint — differential splicing: Show coordinates and strand with the event name. 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. Inspect functional annotations
Gene, transcript and protein annotations help turn a statistical pattern into testable questions. Database matches are context, not experimental observation in the analysed sample.
Working checkpoint — dimensionality reduction: Name databases, versions and prediction status. 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. Control multiple exploration families
Differential tests, correlations and survival scans can each generate many hypotheses. A correction within one analysis does not cover every later click and regrouping.
Working checkpoint — survival association: Keep an exploration log and separate confirmatory families. 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. Audit one event end to end
Trace a selected event from annotation and raw junction counts through PSI, sample grouping, statistical result, plot and external annotation. This catches identifier and denominator errors hidden by dashboards.
Working checkpoint — percent spliced in: Save the complete audit as a worked example. 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. Test group sensitivity
Alternative but reasonable definitions of stage, tissue or expression group can change results. Sensitivity analysis should reveal that dependence rather than search for the most dramatic split.
Working checkpoint — splice-junction count: Compare planned group definitions and retain all outcomes. 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. Test sample influence
Leaving out one independent subject at a time can reveal fragile correlations, components or survival splits. The exercise is diagnostic and cannot replace an external cohort.
Working checkpoint — sample attribute: Flag findings driven by one subject. 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. Compare with vast-tools fairly
vast-tools specialises in read-based event profiling using curated species libraries, while psichomics integrates junction-based splicing with expression and sample attributes. Shared PSI labels may use different event catalogues.
Working checkpoint — differential splicing: Align event paths and sample definitions first. 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. Compare with SpliceWiz fairly
SpliceWiz processes BAM files and stores coverage for event analysis, whereas psichomics can begin from junction counts and curated datasets. Their input evidence and exploratory surfaces differ.
Working checkpoint — dimensionality reduction: Compare the same samples and junction components. 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 school statistics
Medians, variance, correlation and survival curves show why one summary never tells the whole story. Students can practise distinguishing a descriptive pattern from a causal claim.
Working checkpoint — survival association: Ask what alternative variable could create the association. 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 biology
Alternative splicing expands transcript diversity and adds a regulatory layer between genes and possible proteins. The event model remains a simplification of cell-specific RNA processing.
Working checkpoint — percent spliced in: Label observation, model and interpretation in different colours. 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 computing
Interactive interfaces depend on reproducible code, data joins, event annotations and cached resources. A saved session or script is necessary if another analyst must recreate the path.
Working checkpoint — splice-junction count: Export parameters and package versions. 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. Connect to pathways
Integrative transcriptomics brings together biology, statistics, computing and clinical-data literacy and can illuminate routes into research and bioinformatics without promising a credential or job.
Working checkpoint — sample attribute: Check current official course information separately. 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 language
A psichomics analysis can support an association between measured splicing or expression and a declared group or outcome. It does not by itself prove regulation, mechanism, prognosis or treatment relevance.
Working checkpoint — differential splicing: Choose the narrowest defensible verb. 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. Did you know?
psichomics can connect a splicing event to sample attributes, gene expression, dimensional patterns and survival within one environment. The more views available, the more important it becomes to distinguish planned tests from exploration.
Working checkpoint — dimensionality reduction: Keep a click-to-claim audit trail. 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 data releases, sample and subject maps, junction counts, event annotation, expression processing, grouping rules, psichomics version, tests, corrections, exploration log, complete results, plots and validation status.
Working checkpoint — survival association: Make the route from source data to claim replayable. 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