eduKateSG · Why Science?
Test how gene expression changes along one or more inferred lineages—while remembering that pseudotime is an ordering, not automatically a stopwatch
Reading routes
Science learning becomes useful when a familiar object or observation is turned into a system of quantities, mechanisms and claim limits. This guide owns one applied evidence-reading job inside eduKateSG’s wider Science estate. It connects naturally to Why Science Single Cell Rna Sequencing Barcodes Transcriptome Heterogeneity Evidence; Why Science Leap Pseudotime Lagged Coexpression Network Evidence; Why Science Scode Linear Ode Pseudotime Regulatory Network Inference; Why Science Mast Hurdle Models Single Cell Differential Expression; Education Hub; Singapore Secondary School Directory; Career Adulthood Hub; Science Learning Hub. It also keeps current school and public claims traceable to visible primary sources: tradeSeq primary study; tradeSeq PubMed record; tradeSeq open full text; Official Bioconductor tradeSeq package; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science syllabus; 2026 MOE G2 Computing syllabus. The sources describe the scientific scope; this article translates that scope into a calm route for Primary Science, PSLE Science, Secondary Science, O-Level Science, STEM exploration, school choices and career pathways without inventing admission or employment outcomes.
tradeSeq, introduced by Van den Berge and colleagues in Nature Communications in 2020, models gene counts along one or more cellular lineages with negative-binomial generalised additive models. It offers tests for association with pseudotime, differences between lineage patterns, start-versus-end changes and other trajectory contrasts. tradeSeq operates downstream of a trajectory-inference method, so its results are conditional on the cells, lineage structure, pseudotime values, weights and smoothing choices supplied to it; they do not turn an inferred ordering into measured time or causal development.
Inside this guide
1–12 · Foundations and models
- 1. Begin with changing cells
- 2. Define pseudotime
- 3. Define a lineage
- 4. Choose the trajectory upstream
- 5. Name the expression question
- 6. Protect biological samples
- 7. Read the 2020 evidence in scope
- 8. Build a trajectory manifest
- 9. Inspect lineage coverage
- 10. Check the root
- 11. Check branch assignments
- 12. Preserve raw counts
13–24 · Evidence, testing and applications
- 13. Choose the number of knots
- 14. Fit gene-wise smoothers
- 15. Practise with a fictional tradeSeq table
- 16. Use associationTest
- 17. Use patternTest
- 18. Use start-versus-end tests
- 19. Use diffEndTest
- 20. Use earlyDETest
- 21. Read fitted curves with uncertainty
- 22. Adjust for many genes and tests
- 23. Test knot sensitivity
- 24. Test trajectory sensitivity
25–36 · Learning, decisions and pathways
- 25. Compare with LEAP
- 26. Compare with SCODE
- 27. Build a trajectory evidence card
- 28. Design measured-time validation
- 29. Connect tradeSeq to school science
- 30. Try a curve-and-order activity
- 31. Connect to mathematics
- 32. Connect to computing and AI literacy
- 33. Connect to school and career pathways
- 34. Create a family evidence habit
- 35. Use precise tradeSeq language
- 36. Finish with an order-to-curve bundle
Section 1 of 36
1. Begin with changing cells
Development, activation and recovery can place cells along a continuum rather than in tidy categories. Gene expression may rise, fall, peak or diverge as cells move through that process.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 2 of 36
2. Define pseudotime
Pseudotime is an inferred ordering or progress coordinate built from molecular similarity and modelling assumptions. It need not equal minutes, hours or chronological age.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 3 of 36
3. Define a lineage
A lineage is a proposed path through cell-state space. Branches represent alternative routes in the model, not direct observation of every cell’s future.
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 4 of 36
4. Choose the trajectory upstream
Use a suitable trajectory-inference method and document its root, topology, distances and uncertainty. tradeSeq receives this structure rather than replacing it.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 5 of 36
5. Name the expression question
Ask whether a gene changes along any lineage, differs between lineage patterns, changes from start to end or diverges near a branch.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 6 of 36
6. Protect biological samples
Cells supply positions along a path, while donors or experiments support generalisation. Check whether each lineage and pseudotime region includes several samples.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
Section 7 of 36
7. Read the 2020 evidence in scope
Van den Berge and colleagues presented tradeSeq for flexible differential-expression analysis along trajectories using negative-binomial generalised additive models and a family of lineage-aware tests.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 8 of 36
8. Build a trajectory manifest
Record sample, condition, cell annotation, lineage weights, pseudotime, root choice, upstream method and quality metrics for every cell.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 9 of 36
9. Inspect lineage coverage
Plot cells by sample and pseudotime. A late lineage containing cells from only one donor cannot cleanly separate progression from donor identity.
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 10 of 36
10. Check the root
Use measured time, established markers or experimental design to justify the starting region. Reverse or alternative roots should appear in sensitivity analysis when uncertain.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 11 of 36
11. Check branch assignments
Soft lineage weights can represent ambiguous cells near shared paths. Hard labels may discard uncertainty that the model could use.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 12 of 36
12. Preserve raw counts
Negative-binomial modelling expects count information with appropriate offsets. Record how library size and any zero-inflation concerns are handled.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
Section 13 of 36
13. Choose the number of knots
Knots determine the flexibility of smooth expression curves. Use diagnostics or data-driven guidance, then test whether central conclusions survive nearby choices.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 14 of 36
14. Fit gene-wise smoothers
For each gene, model expected counts as smooth functions of lineage-specific pseudotime while sharing information where lineages overlap.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 15 of 36
15. Practise with a fictional tradeSeq table
This teaching table is invented and is not a result from the tradeSeq paper.
| Fictional gene | Association | Pattern difference | Start–end | Reading |
|---|---|---|---|---|
| GENE-R | Strong | Weak | Strong | Shared rise |
| GENE-B | Strong | Strong | Mixed | Branch divergence |
| GENE-F | Weak | Weak | Weak | No supported trend |
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 16 of 36
16. Use associationTest
This test asks whether expression varies with pseudotime along one or more lineages. It does not by itself say where the change occurs or whether lineages differ.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 17 of 36
17. Use patternTest
This contrast asks whether smooth expression patterns differ between lineages over a comparable pseudotime range. Interpret only where both lineages have data support.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 18 of 36
18. Use start-versus-end tests
A start-versus-end contrast summarises change between chosen ends of a lineage. It can miss transient peaks or dips in the middle.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
Section 19 of 36
19. Use diffEndTest
End-point comparisons ask whether lineages finish at different fitted expression levels. The conclusion depends on meaningful, sufficiently sampled lineage endpoints.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 20 of 36
20. Use earlyDETest
Local contrasts near branch regions can search for early divergence. Branch-location uncertainty and sparse cells make precise timing claims especially fragile.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 21 of 36
21. Read fitted curves with uncertainty
Plot observed counts, smooth estimates and uncertainty bands. A tiny adjusted value should not hide a negligible, wildly uncertain or sample-specific effect.
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 22 of 36
22. Adjust for many genes and tests
Declare the gene universe, chosen contrast family and false-discovery procedure. Running every test and reporting only the most exciting result inflates confidence.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 23 of 36
23. Test knot sensitivity
Refit key genes with nearby knot counts. Stable broad trends inspire more confidence than peaks appearing only under one flexible specification.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 24 of 36
24. Test trajectory sensitivity
Repeat the analysis with alternative roots, topology parameters or credible upstream methods. A biological claim should not depend on one arbitrary path.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
Section 25 of 36
25. Compare with LEAP
LEAP searches lagged coexpression along pseudotime. tradeSeq models each gene’s trajectory and tests explicit smooth-curve contrasts across one or more lineages.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 26 of 36
26. Compare with SCODE
SCODE uses linear ordinary differential equations to infer regulatory dynamics from pseudotime. tradeSeq focuses on flexible expression trajectories without claiming a regulatory network.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 27 of 36
27. Build a trajectory evidence card
Record upstream method, root, lineage weights, knots, test and contrast, curve, uncertainty, sample coverage, sensitivities and validation status.
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 28 of 36
28. Design measured-time validation
Collect a real time course, lineage tracing, perturbation or targeted marker panel to test whether the proposed ordering and expression change occur as predicted.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 29 of 36
29. Connect tradeSeq to school science
A seedling’s height can follow a smooth curve over measured days; single-cell pseudotime is different because the order is inferred from many cells rather than tracking one cell.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 30 of 36
30. Try a curve-and-order activity
Arrange shuffled snapshots of a growing plant using visible features, fit a smooth trend and compare the inferred order with the hidden dates.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
Section 31 of 36
31. Connect to mathematics
Smooth functions, count distributions, knots, contrasts and uncertainty show how flexible curves can be tested without confusing fit with mechanism.
tradeSeq fits negative-binomial generalised additive models so gene counts can follow flexible smooth patterns along one or more lineages.
Archive sample manifests, preprocessing decisions, model formulae, contrasts, random seeds, package versions and exclusions so another analyst can reconstruct the decision path.
Section 32 of 36
32. Connect to computing and AI literacy
An elegant trajectory plot is a model output. Its root, geometry, smoothing and data coverage are choices that deserve the same scrutiny as a prediction model.
Pseudotime and lineage weights usually come from another trajectory method; tradeSeq tests expression conditional on those upstream estimates rather than discovering developmental truth independently.
Keep the biological sample as the unit of replication. More cells can sharpen cellular description, but they do not manufacture more independent donors, animals or cultures.
Section 33 of 36
33. Connect to school and career pathways
Students who enjoy changing systems can explore biology, mathematics, computing, biomedical engineering, statistics or developmental science through different routes.
Association, pattern, start-versus-end, end-point and early-divergence tests answer different contrasts, so the selected test must match the written biological question.
Use negative controls, simulated nulls or label permutations only when their exchangeability assumptions match the design, and state exactly what false signal each check can reveal.
Section 34 of 36
34. Create a family evidence habit
When a report says cells progress along a path, ask whether time was measured or inferred, how the root was chosen and whether several samples cover the path.
Knots control smooth-curve flexibility. Too few can miss structure; too many can follow noise, so diagnostic and sensitivity checks are part of the evidence.
Report effect direction, uncertainty, sample support and sensitivity beside adjusted evidence; a ranked gene list or colourful curve is not a complete result.
Section 35 of 36
35. Use precise tradeSeq language
Say expression is associated with inferred pseudotime or differs between fitted lineage patterns. Avoid developmental clock, fate or causal regulator unless additional evidence supports it.
A significant trajectory-linked curve shows statistical association with an inferred ordering. It does not prove elapsed time, lineage commitment or gene regulation.
Create an evidence card naming the comparison, measured modality, statistical unit, model, result, validation status, unresolved alternatives and narrowest defensible claim.
Section 36 of 36
36. Finish with an order-to-curve bundle
Deliver the cell manifest, trajectory alternatives, lineage weights, knot diagnostics, fitted curves, contrasts, sample coverage, multiplicity control, sensitivity checks and measured-time validation plan.
Branch imbalance, uncertain roots and uneven cell density influence fitted curves. Plot lineage weights, cell counts and specimen coverage alongside expression.
Write association, distributional difference or trajectory-linked pattern when that is what was tested. Reserve cause, diagnosis and benefit for designs that support those stronger claims.
## Source date and scope note Primary papers, open records and official software documentation were checked for this article on 10 October 2026. Software interfaces and recommendations can change, so readers should consult the linked current documentation before reproducing an analysis.
## Final reader checklist Before accepting a claim, confirm the biological sample count, condition definition, measured modality, preprocessing, statistical unit, model formula, uncertainty, software version, negative controls, sensitivity checks, independent validation and the exact boundary between association and causation.
Contents · Previous section · Continue to the Science Learning Hub