eduKateSG · Why Science?
Keep every sequencing read beside its quality evidence—then validate the encoding, pairing and laboratory context before drawing conclusions
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. FASTQ is a text format for sequencing reads that pairs each read sequence with an equally long string of encoded quality values. A record begins with an @ identifier line, includes sequence text, uses a + separator, and then carries quality characters corresponding position by position to the sequence. Repository and instrument conventions determine identifier details and quality encoding. A structurally valid FASTQ file does not prove that reads came from the stated sample, that quality values are calibrated, that adapters were removed correctly or that the experiment can answer the biological question. 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: SAM and human-readable sequence alignments; BGZF and blocked compression for genomic data; VCF and structured variant call records; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 12 October 2026: NCBI Sequence Read Archive file-format guide for FASTQ submissions; Official HTSlib faidx manual for indexable FASTQ files; NCBI Sequence Read Archive submission documentation; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Specifications, tools and recommendations can change, so preserve the dated documentation, exact source bytes, parser rules and software versions used.
A deeper reading route
FASTQ joins measurement and uncertainty in the same record. The sequence line reports called bases; the quality line encodes a score for each call. That pairing is powerful because downstream software can trim, filter or weight evidence by position. It also creates a strict invariant: sequence and quality lengths must match after all wrapping conventions are interpreted.
Quality characters are codes, not literal probabilities. A decoder maps each character to an integer according to an agreed ASCII offset, commonly a Phred-style convention in modern data. The score is then related logarithmically to an estimated error probability. Historic encodings differed, so guessing from a filename can turn plausible characters into systematically wrong numbers.
Identifiers carry run and read context but are not universal laboratory provenance. Instruments and converters can encode flow-cell, lane, tile, coordinates, pair number, filtering flags, index sequences or descriptions in different layouts. Pairing should follow the producer contract and manifest rather than a fragile string split copied from another dataset.
Preprocessing changes the scientific object. Adapter trimming, quality trimming, read merging, host filtering, deduplication and error correction can all be useful. Each step changes sequences, quality strings, counts or pairing relationships. A responsible workflow preserves raw data or immutable identifiers, records commands and versions, and reports exactly which file fed each analysis.
FASTQ compression improves storage and transfer but does not repair malformed records or validate quality. Ordinary gzip is common for sequential work; BGZF and indexes support other access patterns under additional rules. Integrity testing should cover the compressed bytes, successful complete decompression and record-level structure.
Quality is one evidence layer. A high base quality says a base caller estimated low error under its model; it does not establish mapping uniqueness, sample identity, variant truth or biological effect. Those questions need alignment, controls, replicate design, contamination checks and downstream statistical evidence.
Quality control works best as a connected dashboard rather than one pass-or-fail number. Per-cycle quality, base composition, adapter content, read length, duplication, contamination and retained-pair counts describe different failure modes. Their interpretation depends on assay design: biased composition may be expected for one protocol and alarming for another. Reports should show distributions, thresholds and denominators, then explain which observations triggered action. A green summary badge without its method is not reproducible evidence.
Sampling uncertainty remains after every base is called. A deeply sequenced region can still be biased by library preparation, amplification or alignment, while a high-quality individual read may not represent the sample population. Replicates, negative and positive controls, coverage breadth and downstream models answer questions that per-base qualities cannot. Keeping these layers separate prevents a technically tidy FASTQ from becoming an overconfident biological story.
A robust handoff includes small known records and expected parser outputs. These fixtures catch accidental changes to wrapping, identifiers, encodings and mate order when software is upgraded. They also make failure understandable: the team can point to the exact invariant that changed instead of rerunning an entire study and hoping that aggregate summaries look similar.
For students, FASTQ turns uncertainty into something visible. Every called letter travels with a confidence code. The uplifting lesson is not that science eliminates uncertainty, but that good data systems preserve it so later decisions can be more careful, transparent and testable.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Capture | Which read and experiment produced the record? | Identifier, manifest and run metadata | A traceable sequencing observation |
| Pair | Does each base have one quality value? | Sequence and quality lengths plus separator rules | A structurally valid record |
| Decode | Which quality convention applies? | Encoding, Phred model and calibration context | Interpretable per-base uncertainty |
| Evaluate | Can the reads support the claim? | QC, controls, preprocessing and downstream validation | A bounded scientific conclusion |
Inside this guide
1–12 · Foundations and design
- 1. Start with the sequencing question
- 2. Recognise the four logical parts
- 3. Read the identifier contract
- 4. Keep identifiers unique enough
- 5. Read the sequence alphabet
- 6. Treat N as uncertainty
- 7. Use the plus separator
- 8. Match sequence and quality lengths
- 9. Decode quality characters
- 10. Understand Phred scaling
- 11. Do not overread calibration
- 12. Inspect quality by cycle
13–24 · Models and estimation
- 13. Inspect base composition by cycle
- 14. Identify adapters explicitly
- 15. Trim with a stated rule
- 16. Filter with a stated denominator
- 17. Handle paired-end reads together
- 18. Distinguish pairing from merging
- 19. Preserve read-group provenance
- 20. Check sample identity
- 21. Assess duplication carefully
- 22. Treat UMIs as separate evidence
- 23. Compress without confusing validation
- 24. Index only when the contract fits
25–36 · Diagnostics and interpretation
- 25. Protect raw data and privacy
- 26. Compare preprocessing versions
- 27. Carry uncertainty downstream
- 28. Connect to Primary Science
- 29. Connect to Secondary Science
- 30. Connect to mathematics
- 31. Connect to computing
- 32. Connect to pathways
- 33. Did you know?
- 34. Write a reader-ready sequencing note
- 35. Build a validation ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Start with the sequencing question
FASTQ records raw or processed reads from a particular assay. Begin with sample, library method, instrument, run and intended downstream use.
Working checkpoint — FASTQ: Write the experiment and file role—raw, trimmed or filtered—before analysis. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 2 of 36
2. Recognise the four logical parts
A conventional record has an @ header, sequence, + separator and quality string. Wrapped variants can span more physical lines.
Working checkpoint — sequencing read: Use a parser that validates logical records rather than assuming every four lines forever. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 3 of 36
3. Read the identifier contract
The text after @ identifies the read under producer-specific conventions and may contain run, coordinate, pairing or filtering fields.
Working checkpoint — quality string: Preserve the complete header and decode it with the correct producer documentation. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 4 of 36
4. Keep identifiers unique enough
Duplicate read names can arise legitimately after concatenation or transformations, but they can also break pairing and provenance assumptions.
Working checkpoint — Phred score: Measure duplicates and document whether uniqueness is required. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 5 of 36
5. Read the sequence alphabet
Read symbols commonly include A, C, G, T and N, while assay and tool rules may permit other symbols or case conventions.
Working checkpoint — paired-end data: Validate every symbol against the declared pipeline contract. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 6 of 36
6. Treat N as uncertainty
N reports an unresolved base call; it is not a fifth DNA base and should not be silently replaced with a convenient nucleotide.
Working checkpoint — read identifier: Count N by cycle and preserve it through transformations unless justified. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 7 of 36
7. Use the plus separator
The + line separates sequence from quality and may repeat the identifier or be otherwise empty after the plus sign.
Working checkpoint — FASTQ: If an identifier is repeated, confirm it matches the record header. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 8 of 36
8. Match sequence and quality lengths
Every sequence position requires one corresponding quality character. A mismatch indicates corruption, wrapping error or faulty transformation.
Working checkpoint — sequencing read: Reject mismatched records before computing any quality statistic. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 9 of 36
9. Decode quality characters
Each quality character maps through an agreed encoding to a numerical score. The same visible character can mean a different value under another offset.
Working checkpoint — quality string: Record the encoding and refuse ambiguous historic data without evidence. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 10 of 36
10. Understand Phred scaling
A Phred-style score expresses estimated error probability on a negative logarithmic scale. Ten score units represent a tenfold change in that estimate.
Working checkpoint — Phred score: Convert a few scores by hand and state the probabilistic model. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 11 of 36
11. Do not overread calibration
A nominal score may be miscalibrated for a chemistry, cycle, instrument or processing stage. Empirical error can differ from the encoded estimate.
Working checkpoint — paired-end data: Use controls or known sites to assess calibration where relevant. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 12 of 36
12. Inspect quality by cycle
Read quality can decline or show instrument-specific patterns across cycles. A single file-wide average can hide local problems.
Working checkpoint — read identifier: Plot quality distributions by position and read group. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 13 of 36
13. Inspect base composition by cycle
Unexpected nucleotide composition may reflect primers, adapters, biased library design, contamination or genuine biology.
Working checkpoint — FASTQ: Compare observed patterns with the assay design before trimming. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 14 of 36
14. Identify adapters explicitly
Adapter sequence can remain when inserts are shorter than reads or trimming was not performed. Adapter evidence depends on the library kit and read layout.
Working checkpoint — sequencing read: Use the declared adapter set and retain trimming reports. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 15 of 36
15. Trim with a stated rule
Quality or adapter trimming changes sequence and quality strings and can change mapping and variant evidence. More trimming is not automatically better.
Working checkpoint — quality string: Report tool, version, thresholds and bases removed. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 16 of 36
16. Filter with a stated denominator
Removing short, low-quality or contaminant reads changes the population of observations. Percent retained needs the original count as denominator.
Working checkpoint — Phred score: Publish counts at every filtering step. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 17 of 36
17. Handle paired-end reads together
Paired reads describe two ends of one library fragment under an expected orientation and distance model. Files and headers must remain synchronised.
Working checkpoint — paired-end data: Check names, order, counts and orphan policy. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 18 of 36
18. Distinguish pairing from merging
Merging overlapping mates creates one consensus read; pairing keeps two observations and their relationship. The choices support different analyses.
Working checkpoint — read identifier: Report how overlaps, disagreements and qualities were resolved. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 19 of 36
19. Preserve read-group provenance
Run, lane, library and sample context help detect batch effects and support downstream grouping. Concatenation can erase distinctions if metadata are not retained.
Working checkpoint — FASTQ: Keep a manifest linking every file and read group to its origin. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 20 of 36
20. Check sample identity
A valid FASTQ can still be swapped, contaminated or labelled incorrectly. File structure cannot authenticate biology.
Working checkpoint — sequencing read: Use appropriate barcodes, controls and independent identity checks. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 21 of 36
21. Assess duplication carefully
Repeated sequences may arise from amplification, high abundance, low complexity or genuine molecules. Duplicate meaning depends on assay and molecular tags.
Working checkpoint — quality string: Avoid one universal duplicate-removal rule. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 22 of 36
22. Treat UMIs as separate evidence
Unique molecular identifiers can help distinguish original molecules from amplification copies when library design and error handling support them.
Working checkpoint — Phred score: Document UMI location, correction and grouping rules. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 23 of 36
23. Compress without confusing validation
Gzip or BGZF reduces storage; it does not validate logical records or quality calibration. Partial decompression can hide a damaged tail.
Working checkpoint — paired-end data: Test complete decompression and record counts. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 24 of 36
24. Index only when the contract fits
HTSlib faidx can index well-formatted FASTQ with sequence and quality wrapping rules, adding a quality offset to .fai. This is not the usual read-pair index.
Working checkpoint — read identifier: Choose indexing only for a defined random-access need. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 25 of 36
25. Protect raw data and privacy
Human sequencing reads can contain identifying and health-related information. Removing names from headers does not anonymise genomes.
Working checkpoint — FASTQ: Apply approved access, encryption and retention controls. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 26 of 36
26. Compare preprocessing versions
Different trimmers or parameters can alter mapping, coverage and calls. Reproducibility needs more than the final FASTQ checksum.
Working checkpoint — sequencing read: Keep commands, reports and representative before-after comparisons. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 27 of 36
27. Carry uncertainty downstream
Base quality can inform alignment and variant models, but later stages add mapping, sampling and model uncertainty.
Working checkpoint — quality string: Do not collapse all uncertainty into one threshold. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 28 of 36
28. Connect to Primary Science
A row of coloured observations paired with confidence symbols shows how uncertainty can travel with data.
Working checkpoint — Phred score: Use invented records and a simple decoding key. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 29 of 36
29. Connect to Secondary Science
FASTQ links measurement, error probability, experimental controls and fair comparison. A confident instrument output can still come from a poor design.
Working checkpoint — paired-end data: Separate technical quality from biological validity. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 30 of 36
30. Connect to mathematics
Logarithms, probabilities, distributions, denominators and paired observations make quality scores rich mathematical examples.
Working checkpoint — read identifier: Convert scores and compare expected error counts. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 31 of 36
31. Connect to computing
Streaming, text encodings, compression, checksums and synchronised files show how systems preserve measurement evidence.
Working checkpoint — FASTQ: Test truncation, malformed separators and length mismatches. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Section 32 of 36
32. Connect to pathways
Sequencing reads connect molecular biology, statistics, software, laboratory operations and data governance and can illuminate STEM pathways.
Working checkpoint — sequencing read: Check current official qualifications separately. Read each quality statistic beside cycle, library method, controls and the exact retained-read denominator.
For future study, connect sequencing to probability, computing and laboratory science while checking pathways separately.
Section 33 of 36
33. Did you know?
FASTQ uses printable characters to store quality values compactly, so one symbol beside a base is a coded number—not a grade or literal probability.
Working checkpoint — quality string: Show the encoding and Phred conversion before interpreting it. Preserve raw reads, filtered-out counts and orphan policy; the final file is not the complete experiment.
For scientific writing, report encoding, preprocessing, denominators, controls and limitations.
Section 34 of 36
34. Write a reader-ready sequencing note
Name sample, assay, instrument, run, read layout, encoding, preprocessing, file hashes, counts and QC limitations.
Working checkpoint — Phred score: Keep raw and processed manifests linked. Separate base-call confidence from mapping confidence, sample identity, variant truth and biological meaning.
For students, ask how one confidence code can travel with every observed symbol.
Section 35 of 36
35. Build a validation ladder
Check compression, complete records, identifiers, alphabet, separators, equal lengths, encoding, pairs, counts and controls.
Working checkpoint — paired-end data: Stop at the earliest unresolved structural or provenance failure. Treat identifiers, quality encoding, pair synchrony and preprocessing versions as scientific inputs.
For parents, ask whether technical quality and biological validity were kept separate.
Section 36 of 36
36. Finish with a reproducible bundle
Deliver immutable hashes, manifests, raw and processed lineage, commands, QC reports, validation logs and bounded downstream claims.
Working checkpoint — read identifier: Make every result traceable to the exact reads used. Keep raw hashes, sample and run manifests, quality encoding, preprocessing commands, counts and QC reports together.
For a classroom model, pair invented letters with a simple confidence key.
Contents · Previous section · Continue to the Science Learning Hub