eduKateSG · Why Science?
A protein record is strongest when sequence identity, review status and evidence remain connected
Start with the accession and organism, distinguish reviewed from unreviewed annotation, then trace individual statements to their evidence and source context.
UniProtKB is a protein knowledgebase, and its records bring sequence information together with functional annotation. The database has reviewed UniProtKB/Swiss-Prot entries and unreviewed UniProtKB/TrEMBL entries. That distinction is useful, but it is not a shortcut for “true” and “false”. A careful reader still checks the organism, sequence, evidence tags, annotation source and the exact statement being used.
This article builds a practical reading routine for students, parents, educators and beginning researchers. It connects secondary science ideas about proteins to real scientific information systems without pretending that a database label replaces experimental evidence. All worked identifiers and protein names in the exercises are fictional unless a source link is explicitly named.
Chapter index
Chapters 1–9: from accession and sequence to evidence tags
- 1. Identify the entry before reading the biology
- 2. Treat the accession as the durable route
- 3. Distinguish reviewed and unreviewed records carefully
- 4. Open evidence tags at the statement level
- 5. Separate protein function from protein existence
- 6. Read the sequence as a specific versioned object
- 7. Interpret sequence features by type and evidence
- 8. Keep isoforms and canonical sequence language visible
- 9. Treat names and synonyms as search aids, not identity proof
Chapters 10–17: from comparisons to responsible reuse
- 10. Use annotation scores as orientation, not proof
- 11. Compare entries one field at a time
- 12. Distinguish missing annotation from negative evidence
- 13. Follow cross-references without merging databases
- 14. Cite the record and the underlying evidence
- 15. Build a protein evidence card
- 16. Use protein records to strengthen classroom explanations
- 17. Connect protein annotation to future pathways
CHAPTER 1 OF 17 · Identify the protein record
1. Identify the entry before reading the biology
Begin with the accession, entry name, protein name, gene name and organism. These fields answer different identity questions. A familiar protein name can occur in several species, and similar gene symbols can hide different records. Copy the accession exactly and keep the organism beside it. This protects every later note about function, sequence or localisation.
Try this
Imagine a fictional protein called Bluefinase with accession `P0BLUE`. One record is from a plant and another similarly named record is from a bacterium. If you copy only “Bluefinase”, a later comparison may mix them. A complete note starts: “P0BLUE, Bluefinase, fictional plant species.”
Create an identity card with accession, reviewed status, organism, sequence length and access date. Do not add gene function yet. The first scientific task is making sure the object is the one you intend to discuss.
Evidence checkpoint
Confirm accession, organism and sequence record before copying a protein name. Similar names are a search convenience, not proof of identity.
CHAPTER 2 OF 17 · Identify the protein record
2. Treat the accession as the durable route
A UniProt accession is designed to identify a knowledgebase record over time. Display names and annotations can change as evidence improves, so use the accession in citations, tables and file names. When comparing two entries, keep both accessions visible even if their protein names appear identical.
Try this
For the fictional record, save notes as `P0BLUE_reading-notes_2026-10-24.txt` rather than `protein-final.txt`. The specific name makes accidental substitution less likely. If the entry later changes, your access date shows which version of the live record you consulted.
Practise with three fictional objects: an accession, a gene symbol and a protein name. Write what each identifies and what it does not. This prevents a gene-level statement from being copied directly into a protein record without checking the relationship.
Evidence checkpoint
Use the accession in filenames and citations, and add the access date. A display name can change while the identifier preserves your route.
CHAPTER 3 OF 17 · Identify the protein record
3. Distinguish reviewed and unreviewed records carefully
UniProtKB/Swiss-Prot is the reviewed section: entries are manually annotated and non-redundant. UniProtKB/TrEMBL is the unreviewed section and contains protein sequences with computationally generated annotation and information from large-scale sources. Review status describes the curation route. It does not mean every reviewed annotation has the same experiment behind it, or that every unreviewed sequence is wrong.
Try this
Suppose fictional entries `P0BLUE` and `A0BLUE` describe related proteins. `P0BLUE` is reviewed; `A0BLUE` is unreviewed. A fair comparison asks which annotations are manually reviewed, which are predicted, and which evidence tags accompany the statements. It does not award a universal quality score to the organisms.
Write two sentences: “This entry is reviewed, which means…” and “To use this function claim, I still need…”. Repeat for an unreviewed entry. The second sentence keeps evidence reading alive after the status label.
Evidence checkpoint
Describe reviewed or unreviewed status as a curation route. Then inspect the evidence attached to the exact function or feature you plan to use.
CHAPTER 4 OF 17 · Identify the protein record
4. Open evidence tags at the statement level
Most UniProtKB information has evidence tags describing its source, such as experimental evidence, similarity, sequence analysis or imported annotation. Evidence belongs to a particular annotation, not automatically to the whole record. Expand the evidence for the statement you plan to use and follow its linked publication or source where available.
Try this
In the fictional record, “binds copper” has an experimental reference, while “located in chloroplast” is predicted from a sequence signal. Both can appear on one page. Writing “Bluefinase was experimentally shown to bind copper and localise to chloroplasts” would merge distinct evidence types.
Make a two-row evidence table with statement, evidence label, source and your conclusion. If a label indicates computational inference, preserve that wording. Accurate evidence language is a core skill in science education.
Evidence checkpoint
Evidence tags belong to annotations, not automatically to the entire entry. Keep experimental, similarity-based, computational and imported statements separate.
CHAPTER 5 OF 17 · Read annotation and evidence
5. Separate protein function from protein existence
Evidence that a protein sequence exists is different from evidence for its biochemical function. A detected peptide may support protein existence without demonstrating a proposed catalytic role. A functional assay may support activity under one condition without mapping every sequence feature. Read the relevant section and its evidence rather than treating “protein exists” as “all annotations are confirmed”.
Try this
Imagine that mass spectrometry detects a peptide unique to fictional Bluefinase. That supports the presence of the protein in the measured material. It does not alone prove copper binding or chloroplast transport. Additional experiments or curated evidence are needed for those claims.
Write three headings—existence, molecular function and biological role—and place each annotation under the best heading. If one statement spans categories, explain why. The exercise reveals how many different questions a single protein page contains.
Evidence checkpoint
Ask whether the evidence supports protein existence, biochemical function, biological role or localisation. Success in one category does not fill the others.
CHAPTER 6 OF 17 · Read annotation and evidence
6. Read the sequence as a specific versioned object
The amino-acid sequence is the foundation of a protein record, but even sequence reading requires context. Check length, sequence status, isoforms or processing where relevant, and any record history that affects comparison. Do not copy a sequence from one organism and annotate it with features from another record because the names look similar.
Try this
For a fictional 100-residue protein, a feature at positions 20–30 covers 11 residues when both endpoints are included. Count carefully. If another entry is 98 residues long, the same position numbers do not guarantee the same biological feature.
Practise by marking a fictional feature from residue 20 to 30 and writing the calculation `30 − 20 + 1 = 11`. Then add the accession beside the diagram. Simple arithmetic becomes reliable only when identity travels with it.
Evidence checkpoint
Write sequence length, isoform and coordinate system beside every residue position. Count inclusive ranges explicitly when a feature boundary matters.
CHAPTER 7 OF 17 · Read annotation and evidence
7. Interpret sequence features by type and evidence
UniProtKB feature annotations can describe domains, active sites, binding sites, modified residues, signal peptides, transmembrane regions and variants. A feature label names the annotation type; its evidence indicates how it was obtained. Predicted topology and experimentally mapped modification are different claims even when both appear as coloured bars.
Try this
In the Bluefinase exercise, residues 1–24 are predicted as a transit peptide, while residue 67 has an experimentally supported modification. Do not say “the entire feature map was experimentally determined”. Report each feature with its own evidence route.
Choose two feature types from a real record and write a sentence for each using the verbs predicts, supports, maps or reports. Avoid the universal verb “proves”. This vocabulary keeps the method visible.
Evidence checkpoint
For each feature, name its type and evidence route. Avoid saying that a full feature map was experimentally determined when some elements are predictions.
CHAPTER 8 OF 17 · Read annotation and evidence
8. Keep isoforms and canonical sequence language visible
Some protein records describe multiple isoforms. A canonical sequence provides a reference representation, but an experiment may concern a different isoform. When a feature or variant uses sequence coordinates, check which sequence is being referenced. A position on one isoform can shift or disappear on another.
Try this
Imagine fictional isoform 1 has 150 residues and isoform 2 lacks residues 51–70. A site at canonical position 90 would correspond to a different numeric position after the deletion if aligned directly. Copying “position 90” into an isoform-2 report without checking the mapping can misidentify the site.
Draw two short sequence bars, mark the deletion and trace one downstream position. The diagram is simple, yet it teaches why accession, isoform and coordinate system belong in the same sentence.
Evidence checkpoint
Check whether a reported position uses the canonical sequence or another isoform. A deletion or alternative segment can shift every downstream coordinate.
CHAPTER 9 OF 17 · Compare entries carefully
9. Treat names and synonyms as search aids, not identity proof
UniProtKB records can include recommended names, alternative names and gene synonyms. These help discovery, but synonyms can be ambiguous across organisms or historical literature. Resolve the record with its accession and taxonomic context before combining claims from different names.
Try this
Suppose “BFA” is used for fictional Bluefinase and also for an unrelated bacterial enzyme. A search for BFA returns both. The synonym helps you find candidates; it does not tell you which paper belongs to which protein. Check organism, sequence and cited identifiers.
Build a synonym table with name, record accession, organism and source. Mark ambiguous abbreviations. This practice connects naturally to vocabulary learning: a term’s meaning depends on the system that defines it.
Evidence checkpoint
Resolve abbreviations through accession and organism before merging papers. One short synonym can refer to unrelated proteins in different contexts.
CHAPTER 10 OF 17 · Compare entries carefully
10. Use annotation scores as orientation, not proof
UniProt annotation scores provide a quick relative indication of how much annotation an entry contains. They are not measures of biological importance, experimental certainty or protein quality. A high score can reflect richer annotation; a low score can reflect limited knowledge. Read the underlying annotations and evidence for any substantive claim.
Try this
Imagine one protein has score 5 and another score 2. You may say the first record has a higher annotation score. You may not conclude that its protein is more important, more active or more evolutionarily conserved. Those are different questions requiring different evidence.
Write three statements about the two scores, then cross out any statement that introduces a concept not contained in the score definition. This is a quick lesson in avoiding metric inflation.
Evidence checkpoint
Use annotation score only as an orientation to record content. Do not convert it into protein importance, activity, certainty or evolutionary conservation.
Comparisons become clearer when you align like with like: organism with organism, sequence with sequence, reviewed status with reviewed status, and evidence-tagged function with the corresponding function. Avoid comparing one entry’s curated summary with another entry’s bare sequence and calling the difference biological.
Try this
For fictional P0BLUE and A0BLUE, construct rows for sequence length, identity after alignment, review status, evidence-backed function and subcellular annotation. Use “not annotated” rather than “absent” when a field is blank. A missing database annotation is not proof that the feature does not exist.
Add a source column to every row. If the comparison draws on another database or paper, cite it separately. The table should reveal where the asymmetry comes from.
Evidence checkpoint
Compare like fields and label missing annotations as not found in the inspected record. Blank space is not biological negative evidence.
CHAPTER 12 OF 17 · Compare entries carefully
12. Distinguish missing annotation from negative evidence
Silence in a record can mean not curated, not predicted, not applicable or not yet known. It rarely supports the strong statement “the protein does not have this property” without an explicit negative annotation or experiment. Database reading therefore requires careful negative language.
Try this
Suppose fictional A0BLUE lacks a localisation entry. The safe note is “no localisation annotation was found in the inspected record on 24 October 2026.” It is unsafe to write “A0BLUE is not localised to chloroplasts”. The first sentence reports your search; the second claims biology.
Practise rewriting three absence claims using “not found in the inspected field”, “not established by this record” and “requires an explicit source”. This habit is valuable far beyond protein databases.
Evidence checkpoint
Rewrite ‘does not have’ as an observation about the searched field unless an explicit negative experiment or curated statement supports absence.
CHAPTER 13 OF 17 · Cite and communicate
13. Follow cross-references without merging databases
UniProtKB records may link to structure, pathway, ontology, genome and proteomics resources. A cross-reference connects records; it does not make every annotation native to UniProtKB or identical across databases. Follow the link, record the destination identifier and describe which resource supports which statement.
Try this
In a fictional workflow, the protein record links to a pathway event and a structure entry. Use UniProtKB for sequence identity, the pathway resource for event context and the structure archive for coordinates. Avoid writing “UniProt experimentally determined the structure” unless the UniProt page explicitly supports that attribution.
Draw a source triangle with accession at the centre and one external identifier at each corner. Label the question answered by each resource. This keeps an integrated workflow honest.
Evidence checkpoint
When following a cross-reference, record the destination resource and identifier. Attribute sequence, pathway and structure statements to the resource that supplies them.
CHAPTER 14 OF 17 · Cite and communicate
14. Cite the record and the underlying evidence
When a claim relies on the knowledgebase record, cite the accession, page and access date. When it relies on a linked experiment, cite the publication as well. A database citation shows your route; a paper citation shows the study. The combination lets another reader inspect both interpretation and source.
Try this
For Bluefinase, a model sentence might be: “The fictional record lists copper binding with experimental evidence; the supporting paper reports an in-vitro assay.” In real work, replace the fictional details with the actual accession, evidence label and citation. Do not cite a homepage for a residue-level claim.
Ask a partner to follow your citation and find the same annotation. If they land on the right protein but cannot locate the evidence, add the section name or evidence code to your note.
Evidence checkpoint
Cite the record for the curated route and the publication for the underlying experiment. A homepage cannot support a residue-level claim.
A compact evidence card makes a long record teachable. Include accession, organism, reviewed status, sequence length, one function statement with evidence, one feature with evidence, and one limitation. The card is a summary of inspected fields, not a replacement for the live record.
Try this
Create a fictional Bluefinase card. Use separate lines for the experimental copper-binding annotation and the predicted transit peptide. Add “No conclusion about in-vivo activity from this exercise.” The limitation is part of the science, not an apology.
Repeat the card for a real school-safe protein topic and compare the two. Your teacher or tutor can then assess whether you preserved identity, evidence and uncertainty, rather than whether you memorised interface labels.
Evidence checkpoint
Limit the evidence card to fields you checked. A concise record with a visible limitation is stronger than a polished card filled from assumptions.
CHAPTER 16 OF 17 · Cite and communicate
16. Use protein records to strengthen classroom explanations
Protein records can turn a textbook statement into a structured inquiry. If the topic is enzymes, inspect active-site annotations and their evidence. If the topic is membranes, inspect transmembrane features and topology. If the topic is inheritance, connect sequence variants to the correct protein coordinates without assuming clinical meaning.
Try this
Choose one curriculum concept and formulate three questions: what sequence object is involved, what annotation describes it, and what evidence supports the annotation? Keep health interpretation outside the exercise unless a qualified, current clinical source is part of the task.
This approach makes science tuition more than extra worksheets. It gives students a calm method for reading live scientific information while respecting the limits of school-level conclusions.
Evidence checkpoint
Choose one curriculum question—enzyme activity, membrane topology or sequence variation—and trace exactly one annotation and its evidence.
CHAPTER 17 OF 17 · Cite and communicate
17. Connect protein annotation to future pathways
UniProtKB reading draws on biology, chemistry, sequence analysis, controlled vocabulary, evidence review and technical writing. Those capabilities appear in biotechnology, proteomics, research software, curation, data science and laboratory work. A student can enjoy one part without committing to an entire career.
Try this
Your next step is to inspect one reviewed and one unreviewed entry for related proteins. Compare identity, evidence tags and one feature. Finish with two sentences: what the records support and what they do not establish. That modest comparison is a strong piece of scientific work.
Continue through the Science Learning Hub, use the Education Hub for school and learning routes, and explore careers by subject capability when you want to connect protein evidence skills to future options.
Evidence checkpoint
Name the capability you enjoyed: sequence reasoning, evidence review, curation or explanation. Treat it as a learning clue, not a promised career outcome.
Contents · Previous chapter · Continue at the Science Learning Hub
A final record-reading routine
Before leaving a UniProtKB entry, copy the accession, organism, review status, sequence length and access date. Then choose only one substantive annotation to trace. Open its evidence tag, identify whether support is experimental, computational, similarity-based or imported, and follow the cited source when the claim matters to your report. This prevents the density of a protein page from turning into an impression that every field has one kind of evidence.
When comparing entries, keep the comparison symmetrical. Put sequence beside sequence, function beside function and feature beside feature. Record ‘not found in the inspected record’ when a field is empty. If one entry is reviewed and the other unreviewed, describe that curation difference without converting it into a judgement about the organism or the protein’s importance.
Finish with a boundary sentence. State what the selected record supports, which evidence route you inspected, and what remains untested. That sentence gives a teacher, tutor or future colleague a clean place to continue the investigation.
Keep the live page link in your notes and revisit it before final submission. Protein knowledgebases are maintained resources, so an access date and a last-minute evidence check are more useful than treating a copied screenshot as permanent. Record any changed annotation and decide whether it alters the sentence you wrote, its evidence boundary, or the comparison itself.
| Reading object | What it can support | Question to keep visible |
|---|---|---|
| Accession and entry | A stable route to a protein knowledge record | Am I using the intended organism and sequence record? |
| Reviewed status | Manual review in UniProtKB/Swiss-Prot | Which statement was reviewed, and what evidence accompanies it? |
| Unreviewed status | Computationally generated or large-scale annotation in UniProtKB/TrEMBL | Which annotations are predictions or imported assertions? |
| Annotation score | A relative indication of annotation content | Am I mistakenly treating this score as experimental proof or biological importance? |
Sources and next reading
Substantive resource claims in this article were checked against the primary pages below. A database record is a structured route into evidence; it does not by itself prove a new biological conclusion. Recheck the live record, evidence labels and release context before using it in assessed work or research.
Primary sources
- UniProt help: UniProtKB — accessed 24 October 2026.
- UniProt help: evidence tags — accessed 24 October 2026.
- UniProt help: querying annotation by evidence — accessed 24 October 2026.
- UniProt help: annotation score — accessed 24 October 2026.
- UniProt help: manual curation — accessed 24 October 2026.