eduKateSG · Why Science?
Find local transcript variants directly from reads—even when a reliable reference genome is missing
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. KisSplice is an exact local transcriptome assembler that can identify SNPs, indels and alternative-splicing events directly from RNA-seq reads with or without a reference genome. It builds a de Bruijn graph, searches for bubble-like alternative paths and quantifies each variant across conditions. Reference-free discovery is powerful for poorly annotated organisms, but graph paths, sequencing errors and biological interpretation still require careful filtering and optional genome mapping. 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: SplAdder and splicing-graph event discovery; Outrigger and de novo single-cell splicing; JunctionSeq and exon–junction usage; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 11 October 2026: Official KisSplice project website; KisSplice primary study full text; KisSplice 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 |
|---|---|---|---|
| Graph | Which k-mers and adjacencies occur in the reads? | Quality-controlled RNA-seq reads and a chosen k | A compressed local transcriptome graph |
| Discover | Which alternative paths form bubbles? | Graph topology and abundance filters | Candidate SNP, indel or splicing variants |
| Quantify | How much read support reaches each path? | Read-to-path evidence across samples | Per-condition variant counts |
| Interpret | What locus and mechanism fit the bubble? | Optional reference mapping and orthogonal data | A bounded local variant hypothesis |
Inside this guide
1–12 · Foundations and design
- 1. Begin when a reference is incomplete
- 2. Use local assembly
- 3. Understand k-mers
- 4. Build the de Bruijn graph
- 5. Compress non-branching paths
- 6. Decompose the graph locally
- 7. Recognise a graph bubble
- 8. Find substitution-like bubbles
- 9. Find indel-like bubbles
- 10. Find splicing-like bubbles
- 11. Name biological replicates
- 12. Control sequencing errors
13–24 · Models and estimation
- 13. Handle repeats and paralogues
- 14. Choose k deliberately
- 15. Quantify each path
- 16. Keep discovery and testing separate
- 17. Use arbitrary conditions carefully
- 18. Map back when a genome exists
- 19. Interpret unmapped bubbles cautiously
- 20. Distinguish allele from isoform
- 21. Inspect path imbalance
- 22. Correct for searched bubbles
- 23. Validate splice boundaries
- 24. Validate with targeted RNA assays
25–36 · Diagnostics and interpretation
- 25. Use long reads for phasing
- 26. Test k-mer sensitivity
- 27. Test sample influence
- 28. Compare with graph methods
- 29. Connect to jigsaw reasoning
- 30. Connect to biodiversity science
- 31. Connect to mathematics
- 32. Connect to computing
- 33. Connect to study pathways
- 34. Use bounded language
- 35. Did you know?
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Begin when a reference is incomplete
Many organisms lack a finished genome or rich transcript annotation. KisSplice can discover local transcript variants from reads directly, so the absence of a reference does not end the scientific question.
Working checkpoint — de Bruijn graph: State whether reference-free analysis is necessary or a deliberate sensitivity check. Keep the independent sample, design, annotation and software version beside this decision so the evidence can be reconstructed.
For a classroom model, use invented counts and clearly label the activity as an analogy rather than a sequencing experiment.
Section 2 of 36
2. Use local assembly
Instead of reconstructing every full transcript, KisSplice assembles local alternatives around variant regions. Local assembly reduces some full-length ambiguity while keeping the claim limited to the reconstructed neighbourhood.
Working checkpoint — k-mer: Describe the assembled interval rather than inventing a whole isoform. 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. Understand k-mers
Reads are decomposed into overlapping words of length k. Shared k-mers connect sequences in a de Bruijn graph, so k controls the balance between collapsing repeats and fragmenting weak paths.
Working checkpoint — graph bubble: Record k and justify it for read length and complexity. 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. Build the de Bruijn graph
Graph nodes or compacted segments represent sequence words and edges represent observed adjacency. True variants and technical errors can both create branches.
Working checkpoint — local assembly: Keep the read-quality and abundance filters with the graph. 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. Compress non-branching paths
Long stretches without a choice can be compacted so the algorithm focuses on branch points. Compression improves tractability without adding biological evidence.
Working checkpoint — alternative splicing: Preserve the mapping from compacted sequence back to supporting k-mers. 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. Decompose the graph locally
Large transcriptome graphs contain many disconnected or weakly connected regions. Local decomposition lets KisSplice search manageable components for variant patterns.
Working checkpoint — reference-free discovery: Report component size and any computational limits. 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. Recognise a graph bubble
A bubble consists of alternative paths that diverge and later rejoin. SNPs, indels and alternative-splicing events can all create bubble-like structures, so topology begins discovery but does not finish classification.
Working checkpoint — de Bruijn graph: Inspect path lengths, sequences and support before naming the variant. 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. Find substitution-like bubbles
A single-nucleotide difference can create short alternative paths. Sequencing error, paralogous sequence and genuine polymorphism are competing explanations.
Working checkpoint — k-mer: Require abundance and consistency beyond one read. 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. Find indel-like bubbles
Small insertions or deletions create paths of different length. Alignment to a trusted reference can help classify them, but the graph itself records only sequence alternatives.
Working checkpoint — graph bubble: Report both path sequences and coverage. 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. Find splicing-like bubbles
An alternative exon or splice connection can create a longer asymmetric bubble. The exact graph geometry depends on read length, k and transcript abundance.
Working checkpoint — local assembly: Use path structure and optional genomic mapping together. 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. Name biological replicates
Counts across conditions support inference only when sample labels reflect independent organisms, cultures or specimens. Merging reads too early can erase replicate variability.
Working checkpoint — alternative splicing: Keep per-sample files and condition metadata. 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. Control sequencing errors
Low-quality bases produce rare k-mers and spurious branches. Error removal improves graph clarity but can also remove genuine low-abundance variants.
Working checkpoint — reference-free discovery: Publish quality and abundance thresholds. 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. Handle repeats and paralogues
Shared sequence among gene families can connect reads from different loci. A locally plausible bubble may therefore combine sources that are separate in the genome.
Working checkpoint — de Bruijn graph: Use mapping or longer reads to challenge ambiguous components. 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. Choose k deliberately
A small k increases connectivity and repeat collapse; a large k demands more exact coverage and can split low-expression paths. There is no universally correct value.
Working checkpoint — k-mer: Test a small set of defensible k values for priority events. 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. Quantify each path
KisSplice counts evidence for alternative paths across conditions. Counts are tied to the assembled sequence and read-assignment rules rather than a pre-existing gene model.
Working checkpoint — graph bubble: Retain path sequences with the count matrix. 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. Keep discovery and testing separate
Finding a bubble defines a candidate; testing whether its path usage differs across conditions is a second question. Reusing the same data is acceptable when the inferential method accounts for the workflow, but the distinction remains important.
Working checkpoint — local assembly: Label discovery filters and comparison rules 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 17 of 36
17. Use arbitrary conditions carefully
The project describes quantification across an arbitrary number of biological conditions. More groups increase modelling choices and multiple comparisons, not automatic explanatory power.
Working checkpoint — alternative splicing: Predefine contrasts and adjustment families. 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. Map back when a genome exists
Reference-free discovery can be followed by mapping assembled paths to a genome or transcriptome. Mapping helps locate genes, splice boundaries and paralogous matches.
Working checkpoint — reference-free discovery: Report all high-quality mapping locations. 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. Interpret unmapped bubbles cautiously
A path may fail to map because the reference is incomplete, the sequence crosses a novel junction, contamination is present or assembly is erroneous. Unmapped does not mean newly discovered biology.
Working checkpoint — de Bruijn graph: Investigate taxonomy, complexity and raw support. 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. Distinguish allele from isoform
SNPs and splice alternatives can both produce graph bubbles. Relative abundance across conditions alone may not tell whether the paths are alleles, paralogues or RNA-processing choices.
Working checkpoint — k-mer: Use path geometry, genomic placement and sample genotypes. Treat identifiers, denominators and factor levels as scientific quantities: a silent mismatch can reverse an otherwise correct calculation.
For parents, a calm checkpoint is to ask how many independent samples support the claim and which assumption matters most.
Section 21 of 36
21. Inspect path imbalance
One path may have much lower coverage because of true rare usage, mapping difficulty or errors. Ratios should be accompanied by absolute counts and sample consistency.
Working checkpoint — graph bubble: Plot both paths for every biological replicate. 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. Correct for searched bubbles
A transcriptome-scale graph can yield many candidates. Filtering and statistical testing define the family in which false discoveries are controlled.
Working checkpoint — local assembly: Keep all tested bubbles and adjusted 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 23 of 36
23. Validate splice boundaries
When genomic sequence is available, canonical splice motifs and boundary alignment support a splicing interpretation. Non-canonical splicing exists, so motif absence is a warning rather than absolute disproof.
Working checkpoint — alternative splicing: Report motif rules and genome release. 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. Validate with targeted RNA assays
Primers designed around alternative paths can test whether both sequence forms occur in independent material. Assay design must avoid regions shared with paralogues.
Working checkpoint — reference-free discovery: Use a new biological sample where possible. 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. Use long reads for phasing
Long-read RNA sequencing can connect the local bubble to more distant transcript structure. It adds phasing information but retains its own error and coverage limits.
Working checkpoint — de Bruijn graph: Match long-read evidence to the same locus and condition. 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 k-mer sensitivity
A candidate that appears only at one fragile k may reflect graph construction more than stable biology. Conversely, real low-coverage variants may disappear under strict settings.
Working checkpoint — k-mer: Describe stability without demanding identical graphs. 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
Leave-one-replicate-out analysis reveals bubbles or usage differences driven by one specimen. The omission should follow biological units, not arbitrary read subsets.
Working checkpoint — graph bubble: Flag discoveries that repeatedly vanish or reverse. 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 graph methods
SplAdder and SGSeq build genome-guided splice graphs from aligned reads, while KisSplice begins with a read-derived de Bruijn graph and can work without a genome. Their graph nodes and evidence units differ fundamentally.
Working checkpoint — local assembly: Compare reconstructed sequences and coordinates, not total events. Preserve the complete tested family and every excluded feature; a polished shortlist is not a substitute for provenance.
For scientific writing, prefer “supports”, “is associated with” and “is predicted to” when stronger causal verbs are not justified.
Section 29 of 36
29. Connect to jigsaw reasoning
A de Bruijn graph resembles solving many overlapping word fragments: repeated fragments can connect the wrong stories, while unique overlaps clarify a path.
Working checkpoint — alternative splicing: Use short invented strings to model k-mer overlap. 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 biodiversity science
Reference-free methods are valuable for non-model organisms whose genomes or annotations remain incomplete. They broaden whose biology can be studied rather than making references irrelevant.
Working checkpoint — reference-free discovery: Pair computational discovery with careful specimen and species records. 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 mathematics
Graphs, paths, branching and cycles turn sequence assembly into discrete mathematics. Parameters such as k change the graph and therefore the set of possible explanations.
Working checkpoint — de Bruijn graph: Ask students to predict how larger k changes connectivity. 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
Memory, graph decomposition and exact algorithms determine whether hundreds of millions of reads can be analysed. Computational feasibility is part of experiment planning.
Working checkpoint — k-mer: Record hardware, runtime and software environment. 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 study pathways
Reference-free transcriptomics brings together genetics, ecology, statistics and algorithms and can illuminate routes into genomics, conservation science or bioinformatics.
Working checkpoint — graph bubble: Avoid turning one tool into a credential promise. 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 KisSplice bubble supports local alternative sequences observed in the read graph. Classification as splicing, polymorphism or indel and any functional claim require additional evidence.
Working checkpoint — local assembly: Make the narrow graph claim first. 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?
KisSplice can analyse RNA-seq with or without a reference genome because it searches for alternative paths directly in a graph built from reads. A reference becomes valuable context, not a prerequisite for every discovery.
Working checkpoint — alternative splicing: Reference-free does not mean evidence-free. 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 read accessions, sample metadata, QC, KisSplice version, k and filters, commands, graph-path sequences, counts, mappings, complete tests, sensitivity checks and validation status.
Working checkpoint — reference-free discovery: A local assembly should remain traceable to the reads that created it. 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