eduKateSG · Why Science?
Let the observed data teach you about technical noise—then ask whether a biological signal is convincingly larger
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 Deseq2 Moderated Dispersion Fold Change Count Data; Why Science Edger Tmm Quasi Likelihood Replicated Counts; Why Science Limma Voom Precision Weights Linear Models Rna Seq; Why Science Dss Dispersion Shrinkage Count Data; 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: NOISeq original article; NOISeq quality-control study; Official Bioconductor NOISeq package; Current official NOISeq guide; 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.
NOISeq is a Bioconductor toolkit for RNA-seq quality control, normalisation, filtering and nonparametric differential-expression analysis. Its differential methods build an empirical noise distribution from the data instead of beginning with a fixed count distribution. The current official guide distinguishes NOISeqBIO for experiments with biological replication from simulation-based options used in other settings. That flexibility is not permission to overlook design: sample independence, count provenance, batch structure, filtering and diagnostics still determine what a probability-like score can support.
Inside this guide
1–12 · Foundations and models
- 1. Begin with the scientific unit
- 2. Separate measurement from biology
- 3. Write the comparison first
- 4. Meet the uncertainty
- 5. Identify the central assumption
- 6. Keep the claim bounded
- 7. Read the primary evidence in scope
- 8. Build a trustworthy input matrix
- 9. Create a sample sheet
- 10. Inspect library composition
- 11. Check sample relationships
- 12. Filter by information
13–24 · Evidence, testing and applications
- 13. Confirm the design is estimable
- 14. Choose normalisation deliberately
- 15. Practise with a fictional evidence table
- 16. Run the method with recorded settings
- 17. Interpret the central statistic
- 18. Correct for many features
- 19. Read effects beside probabilities
- 20. Plot every biological sample
- 21. Inspect the method-specific diagnostic
- 22. Challenge the key assumption
- 23. Test filtering sensitivity
- 24. Test normalisation sensitivity
25–36 · Learning, decisions and pathways
- 25. Test sample influence
- 26. Test parameter sensitivity
- 27. Compare an adjacent method
- 28. Build an evidence card
- 29. Connect to fair testing
- 30. Try a classroom activity
- 31. Connect to mathematics
- 32. Connect to computing literacy
- 33. Connect to education pathways
- 34. Create a family question habit
- 35. Use precise scientific language
- 36. Finish with an evidence bundle
Section 1 of 36
1. Begin with the scientific unit
Define the independent sample before reading any count table; this prevents depth, lanes or simulated replicates from being mistaken for biological replication.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 2 of 36
2. Separate measurement from biology
A gene count is a computational summary of sequenced fragments under an annotation and pipeline, not a direct census of molecules in a cell.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 3 of 36
3. Write the comparison first
Name the condition, tissue, time point and direction of interest. A method cannot repair an unclear or post-hoc scientific question.
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 4 of 36
4. Meet the uncertainty
Biological variation and technical processing create within-group change. NOISeq asks whether between-group change stands out from that empirical background.
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 5 of 36
5. Identify the central assumption
The observed within-condition comparisons must reasonably represent the noise relevant to the between-condition question; outliers and hidden batches challenge that premise.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 6 of 36
6. Keep the claim bounded
A feature ranked by NOISeq is evidence for a measured abundance difference under this dataset and workflow, not proof of regulation or function.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
Section 7 of 36
7. Read the primary evidence in scope
Tarazona and colleagues introduced NOISeq as a nonparametric approach, while later work expanded systematic quality control. Read performance claims within those evaluated data.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 8 of 36
8. Build a trustworthy input matrix
Use documented integer feature counts with columns matched exactly to samples. Record alignment or quantification, counting rules, annotation release and strandedness.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 9 of 36
9. Create a sample sheet
Store condition, batch, subject, sex, time and processing covariates beside each sample. Check duplicates, spelling and missing values before analysis.
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 10 of 36
10. Inspect library composition
Examine totals, dominant features, detected-gene counts and composition. One extreme library can reshape relative abundance across many rows.
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 11 of 36
11. Check sample relationships
Use sample correlations and ordination to find unexpected grouping, but investigate rather than automatically delete a sample that looks unusual.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 12 of 36
12. Filter by information
Apply a declared information rule before modelling. Save the retained universe because filtering changes both noise estimation and the multiple-testing family.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
Section 13 of 36
13. Confirm the design is estimable
If condition is perfectly confounded with batch, no software can attribute their shared pattern uniquely. Redesign or narrow the inference.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 14 of 36
14. Choose normalisation deliberately
Compare defensible normalisation procedures using diagnostics and biological context. Record factors and avoid choosing the method that maximises discoveries.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 15 of 36
15. Practise with a fictional evidence table
This classroom table is invented and is not a result from a NOISeq study.
| Fictional gene | M | D | NOISeq probability | Reading |
|---|---|---|---|---|
| GENE-N1 | 1.4 | 82 | 0.96 | Follow up |
| GENE-N2 | -0.5 | 9 | 0.61 | Weak separation |
| GENE-N3 | 2.1 | 3 | 0.88 | Check low counts |
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 16 of 36
16. Run the method with recorded settings
Use the appropriate NOISeq function for the replication structure, keep all arguments, and save the software version and random seed where simulation is involved.
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 17 of 36
17. Interpret the central statistic
Read probability, M and D together. A high method-specific probability needs an effect with sensible abundance and sample agreement.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 18 of 36
18. Correct for many features
State the probability or false-discovery rule used across all tested features and comparisons. Do not retrofit a threshold around attractive genes.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
Section 19 of 36
19. Read effects beside probabilities
Magnitude, direction and uncertainty answer different questions. Report them together so small but stable and large but noisy changes remain distinguishable.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 20 of 36
20. Plot every biological sample
Show the individual normalised counts behind every headline. Group summaries can hide one library driving the apparent difference.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 21 of 36
21. Inspect the method-specific diagnostic
Review NOISeq quality plots for saturation, count distribution, composition and feature-type behaviour. Each diagnostic addresses a different possible failure.
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 22 of 36
22. Challenge the key assumption
Ask whether within-group differences are exchangeable with the technical and biological noise relevant to the contrast, especially when groups have unequal variability.
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 23 of 36
23. Test filtering sensitivity
Repeat the workflow across a small set of justified count thresholds. Results that depend on one arbitrary cutoff need cautious language.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 24 of 36
24. Test normalisation sensitivity
Compare key results under another defensible scaling rule. Composition-sensitive reversals should be described rather than hidden.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
Section 25 of 36
25. Test sample influence
When sample size allows, leave out one biological sample at a time and record probability, sign and rank changes.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 26 of 36
26. Test parameter sensitivity
Vary justified NOISeq settings without chasing a desired list. Document which features remain stable across plausible choices.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 27 of 36
27. Compare an adjacent method
Compare with DESeq2, edgeR or voom only after aligning count universe, design and contrast; disagreements can expose assumptions worth investigating.
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 28 of 36
28. Build an evidence card
Record input provenance, replication route, normalisation, filters, M, D, probability, plots, sensitivities, versions and planned validation.
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 29 of 36
29. Connect to fair testing
The logic mirrors a fair school experiment: estimate ordinary variation under controlled conditions before deciding whether a treatment difference is unusual.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 30 of 36
30. Try a classroom activity
Let students measure repeated background readings, create an empirical difference distribution, then test whether a labelled signal exceeds that noise.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
Section 31 of 36
31. Connect to mathematics
Ratios, absolute differences, resampling, ranks and distributions turn the everyday idea of signal versus noise into explicit evidence.
NOISeq constructs an empirical distribution of within-condition changes and compares between-condition change against that observed noise; the evidence is data-adaptive rather than distribution-free magic.
Archive raw-data accessions, sample metadata, annotations, software versions, commands, filters and random seeds so another analyst can reconstruct every decision.
Section 32 of 36
32. Connect to computing literacy
A reproducible script makes filters, factor levels and seeds inspectable. Computing choices are part of the experiment, not invisible administration.
NOISeqBIO is the biologically replicated route described in the official guide. Technical replication or simulated pseudo-replicates cannot substitute for independent sampling.
Count independently collected organisms, people, cultures or specimens as the biological replicates; computational resamples and repeated reads do not create new experimental units.
Section 33 of 36
33. Connect to education pathways
RNA-seq joins biology, statistics and computing, giving learners routes through life science, data science, biotechnology and research careers.
The M statistic represents log fold change and D represents absolute expression difference; using both helps distinguish large ratios at tiny counts from substantial abundance shifts.
Pair adjusted evidence with effect direction, uncertainty, sample-level plots, diagnostics and at least one sensitivity analysis capable of contradicting the headline.
Section 34 of 36
34. Create a family question habit
When a gene is called different, ask: different compared with what observed noise, across how many independent samples, and by how much?
A NOISeq probability is method-specific evidence that change exceeds estimated noise. It is not automatically a conventional p-value, posterior probability of truth or mechanism score.
Declare the testing family, comparisons and exclusions before selecting examples. Multiple testing belongs to the complete decision set, not merely the displayed rows.
Section 35 of 36
35. Use precise scientific language
Say that NOISeq found a measured expression difference exceeding its empirical noise model under the stated settings. Avoid claiming causation.
Quality-control modules examine counts, saturation, composition, length bias and biotypes. A completed differential table does not erase problems identified upstream.
Build an evidence card naming the question, measured material, statistical unit, model, contrast, result, alternatives, validation status and narrowest defensible claim.
Section 36 of 36
36. Finish with an evidence bundle
Deliver counts, metadata, quality reports, code, settings, complete results, sample plots, sensitivity analyses, versions and the bounded interpretation.
Nonparametric modelling reduces dependence on a named count distribution, but it still depends on exchangeability, appropriate normalisation, filtering and representative replicates.
Use association, differential abundance or differential expression when that is what was estimated; reserve cause, mechanism, diagnosis and benefit for stronger evidence.
## Source date and scope note Primary studies and current official software documentation were checked for this article on 11 October 2026. Package interfaces, recommendations and dependencies can change; readers should consult the linked official package page and vignette before reproduction.
## Final reader checklist Before accepting a claim, confirm the biological sample count, condition definition, measured material, annotation, preprocessing, statistical unit, exact comparison, effect size, uncertainty, software versions, controls, sensitivity checks, independent validation and the boundary between association and causation.
Contents · Previous section · Continue to the Science Learning Hub