# Understanding UniProt Entry Pages

A UniProt entry page is the structured, curated web record for a single [protein sequence](/blog/guides/protein-sequence) in the UniProt Knowledgebase. It gathers function, names, taxonomy, subcellular location, sequence, variants, interactions, structures, and external links into one document, with the supporting evidence attached to nearly every individual claim.

If you have ever tried to answer a simple question about a protein, such as which domains it contains, whether it is found in the nucleus, or which papers reported its phosphorylation sites, you already understand why this page matters. Protein information lives in dozens of specialized databases, each with its own formats and search conventions, and manual searching across them is slow, error prone, and often incomplete. The UniProt entry page solves this by pulling that distributed information into a single, standardized record that you can read, filter, download, and cite. Learning to read it well is one of the highest-return skills in practical bioinformatics, because the same reading habits transfer to every protein you will ever analyze.

This tutorial is written for students and working scientists who want to use UniProt confidently, not just occasionally. We will walk through the entry page section by section, decode the evidence and quality indicators, work through a concrete example, and cover the workflows where UniProt earns its keep.

## What Is a UniProt Entry Page, Exactly?

The UniProt Knowledgebase, usually abbreviated UniProtKB, contains two distinct classes of records, and the difference shapes everything you see on the page.

**Reviewed entries** (the Swiss-Prot component) are annotated by professional curators who read the primary literature, verify claims, and attach explicit evidence to each statement. Reviewed entries have full annotation of function, domains, post-translational modifications, and disease links where the data exist.

**Unreviewed entries** (the TrEMBL component) are annotated automatically, largely by rule-based transfer from reviewed proteins and by automated pipelines. TrEMBL provides breadth, covering hundreds of millions of sequences from genome projects and metagenomes, while Swiss-Prot provides depth, covering roughly half a million well-characterized entries.

When you search UniProt, results from both classes appear together, and each result row tells you which class it belongs to. A reviewed entry carries a gold "Reviewed" badge. An unreviewed entry carries a silver "Unreviewed" badge. Everything else on the page, from the annotation style to the number of literature citations, follows from this distinction.

### The header of the entry page

The top of every entry page carries the identity information that you will use most often:

- **Protein recommended name**, such as "Cellular tumor antigen p53".
- **Entry name**, such as P53_HUMAN. This follows a convention of a short protein designation, an underscore, and an organism code (HUMAN, MOUSE, ECOLI).
- **Accession number**, such as P04637. This is the stable identifier.
- **Organism and gene name(s)**, for example Homo sapiens and TP53.
- **Sequence length and mass**, for example 393 amino acids and roughly 43.7 kDa.
- **Protein existence (PE) level**, a number from 1 to 5 that states how well the existence of the protein is supported.
- **Annotation score**, a value from 1 to 5 that summarizes how much annotation the entry carries.
- Action buttons for downloading the record, running BLAST, aligning sequences, and mapping identifiers.

Two identifiers appear in the header and students routinely confuse them. The **accession** is permanent. Once assigned, it is never reused and never changed, which is why publications and databases cite accessions. The **entry name** is a human-readable label, and UniProt can replace it when nomenclature improves, keeping the old name listed among previous names. If you want reproducibility, cite the accession. If you want readability in a figure or a talk, use the entry name.

Protein existence levels deserve a moment of attention, because they set expectations for the whole record:

1. **PE 1, evidence at the protein level.** The protein has been observed experimentally, for example by mass spectrometry or purification.
2. **PE 2, evidence at the transcript level.** A full-length mRNA has been sequenced.
3. **PE 3, inferred by homology.** The protein exists by similarity to related proteins.
4. **PE 4, predicted.** The sequence comes from gene prediction alone.
5. **PE 5, uncertain.** The existence of the protein is doubtful.

A PE 1 reviewed entry and a PE 4 unreviewed entry can sit next to each other in search results, yet they represent very different levels of confidence. Reading the PE level first is the fastest triage step on any entry page.

## Summary Table: Key Terms on a UniProt Entry Page

| Term | What it means | Term it is most often confused with | Key difference |
|--|--|--|--|
| Accession (e.g., P04637) | Stable, permanent identifier for the entry | Entry name (e.g., P53_HUMAN) | Accessions never change, entry names can be replaced |
| Reviewed (Swiss-Prot) | Manually curated by UniProt biologists | Unreviewed (TrEMBL) | Reviewed entries carry literature-verified annotation |
| Canonical sequence | The reference sequence of the entry | Isoform sequence | Isoforms arise from [alternative splicing](/blog/guides/alternative-splicing) or initiation and are listed as variants of the canonical sequence |
| Entry version | Version of the annotation content | Sequence version | A change in annotation bumps the entry version, a change to the [amino acid sequence](/blog/guides/amino-acid-sequence) bumps the sequence version |
| UniProtKB | Knowledgebase with function annotation | UniParc | UniParc is a non-redundant archive of all sequences without annotation |
| UniRef clusters | Sequence sets clustered at 100%, 90%, and 50% identity | UniProtKB entries | UniRef compresses redundancy for faster searches and broad coverage |
| Experimental evidence | Backed by a publication or direct assay | Predicted or inferred evidence | Evidence tags on each annotation show which is which |
| AlphaFold model | Computationally predicted structure | PDB cross-reference | PDB entries are experimentally determined structures, AlphaFold entries are predictions |

## UniProt Entry Page Tabs Explained: The Sections of an Entry

UniProt divides each entry into sections, reached from the navigation menu on the left side of the page. In the current interface these function as tabs, and the grouped layout follows the same organizational logic that partner resources have adopted. The AlphaFold [Protein Structure](/blog/guides/protein-structure) Database, for example, recently redesigned its own entry pages around dedicated summary and domains tabs with annotations integrated alongside an interactive 3D viewer, aligned to a UniProt release, which reflects how the wider protein-informatics ecosystem now organizes entry content [1]. The following walkthrough covers each section you will encounter, in roughly the order UniProt presents them.

### Names & Taxonomy

This section lists the recommended protein name, alternative names given in the literature, the EC number if the protein is an enzyme, the gene names with their official symbol, and the full taxonomic lineage drawn from the NCBI taxonomy. The taxon identifier is a clickable link into UniProt's taxonomy browser, where you can see every entry UniProt carries for that organism.

Practical reading tips:

- The recommended name is the name curators consider standard. Alternative names are preserved because older papers use them.
- Gene names follow organism-specific nomenclature committees. For human genes, the HGNC symbol is authoritative.
- The lineage matters when you are choosing an ortholog or deciding whether a statement about "mouse" applies to "human."

### Subcellular location

Here UniProt records where the protein localizes in the cell, using a controlled vocabulary of compartment terms such as nucleus, cytoplasm, mitochondrion, and plasma membrane. For membrane proteins, the **topology** sub-entries describe which segments cross the membrane and which side each region faces, and the **transmembrane** features appear in the sequence viewer as labeled helices.

Two kinds of claims appear here. Some localization statements come from experiments and carry citations. Others are predictions by computational methods such as those that assign subcellular location by sequence similarity or signal-based inference, and UniProt marks those as potential or predicted. Always check the qualifier word next to the location. "Located in the nucleus" and "potential nuclear localization" are different claims with different weight.

### Function

The Function section is often the first stop for interpretation, and for a reviewed entry it is dense and high-value. It contains:

- A plain-language functional summary written by curators.
- **Enzyme specificity** for enzymes, including reactions, cofactors, and regulation.
- **Keywords**, a controlled vocabulary of one- to three-word terms (for example, "DNA-binding", "ATP-binding", "Apoptosis") shown as clickable tags that lead to all entries sharing the term.
- **Gene Ontology (GO) annotations** in three categories: molecular function, biological process, and cellular component.
- Pathways, cross-linked to resources such as Reactome.

Each GO annotation carries an **evidence code**. Codes beginning with "IEA" mean inferred by electronic annotation, which is purely computational. Codes such as "IDA" (inferred from direct assay) or "IMP" (inferred from mutant phenotype) mean a specific experiment backs the claim. When you write "the protein is involved in DNA repair" in a manuscript or a report, the strength of that sentence depends entirely on whether the supporting evidence is experimental or electronic, and the Function section is where you check.

### Pathology & Biotech

For proteins linked to human disease, this section lists the associated conditions, usually with cross-references to OMIM (Online [Mendelian Inheritance](/knowledge/molecular-biology/mendel-heredity-laws-of-inheritance-explained-simply) in Man). It also records **mutagenesis** experiments, where researchers mutated a specific residue and observed a defined effect, and **pharmaceutical** information where a protein is a drug target.

This section rewards careful reading of feature positions. If you are studying a patient variant or a CRISPR mutation, you can look up the exact residue here, see whether other mutations at that position have documented consequences, and find the citation describing the experiment. For many clinically important proteins, this is the single fastest route from a variant coordinate to a body of prior experimental work.

### PTM / Processing

Post-translational modification and processing annotations live here:

- Signal peptides, transit peptides, and propeptides that are removed during maturation.
- Chain and peptide assignments after cleavage.
- Modification features such as phosphorylation, acetylation, ubiquitination, and glycosylation, each with the modified residue, the modifying enzyme where known, and the reference.
- Lipid anchors and cross-links.

The practical value is position-level precision. A phosphorylation entry does not just say "the protein is phosphorylated." It says which serine, threonine, or tyrosine, sometimes with multiple independent studies cited, and sometimes with the upstream kinase named. If you are designing a phospho-specific antibody, interpreting a mass-spectrometry hit, or reasoning about a signaling cascade, this section supplies the residue-level facts.

### Expression

This section describes where and when the protein is expressed, using evidence from sources such as tissue-specific transcript and protein studies. For human proteins, UniProt cross-references tissue expression resources. Treat expression claims with the same evidence-awareness as localization claims, because developmental stage, tissue, and assay type all change what a statement means.

### Interaction

The Interaction section reports binary protein-protein interactions, sourced largely from the IntAct database. Each interaction lists the partner protein, an optional experimental description, and the number of publications supporting it. UniProt distinguishes direct physical interactions from those observed in complexes.

A useful habit is to sort or scan interactors by how many distinct publications support them. An interaction backed by five independent studies deserves different treatment in your model than one reported in a single high-throughput screen, and UniProt gives you the basis to make that call.

### Structure

This section connects the sequence to three-dimensional structure in two distinct ways.

**Experimental structures** appear as cross-references to the [Protein Data Bank](/blog/guides/protein-data-bank) (PDB). Each PDB entry tells you the method (X-ray crystallography, cryo-electron microscopy, NMR), the resolution, and which segment of the protein the structure covers. Coverage is the detail students miss most often. A structure that resolves residues 94 to 312 tells you nothing directly about the conformation of the first 93 residues, and the Structure section makes those gaps visible.

**Predicted structures** appear through the AlphaFold Protein Structure Database. AFDB, developed by EMBL-EBI and Google DeepMind, provides open access to hundreds of millions of predicted structures, has become integrated into major databases and analysis platforms, and is aligned to UniProt releases so that each predicted model corresponds to a specific UniProt sequence state [1]. UniProt links to the AlphaFold model directly from the entry, and the model loads in an interactive viewer where you can rotate the chain, highlight domains, and map features onto the folded structure. The distinction to hold in mind is simple. PDB entries are measurements. AlphaFold entries are predictions with per-residue confidence values, and both are useful in different ways.

A third element here is the **3D viewer** embedded in UniProt itself, which lets you inspect PDB models without leaving the page. For quick questions such as "is my residue buried or surface-exposed," the embedded viewer is often enough.

### Family & Domains

This section describes the protein's domain architecture and its position in protein families:

- **Domains** from signature databases such as Pfam, PROSITE, and SMART, each with residue boundaries and a link to the source database.
- **Regions and repeats** annotated by curators or detected computationally.
- **Motifs**, short conserved sequences such as zinc finger signatures or nuclear localization signals.
- **Compositional bias** such as low-complexity or acidic regions.
- **Family membership**, placing the protein in a hierarchy of related families.
- **Sequence similarities** to related proteins, often grouped so you can see paralogs within the same species and orthologs across species.

Domain boundaries are among the most-used pieces of information in all of bioinformatics, because truncation constructs for expression, mutation interpretation, and homology modeling all depend on knowing where a domain starts and ends. When a domain annotation includes a note such as "by similarity," it was transferred from a homolog rather than demonstrated in this protein.

### Sequence

The Sequence section contains the canonical amino acid sequence in full, its molecular mass and length, and the **isoforms** produced by alternative splicing or alternative initiation. Each isoform is described by the differences from the canonical sequence (for example, "missing residues 1 to 45" or "contains an alternate in-frame exon"), and each isoform carries its own stable identifiers.

Down in this section you will also find the **sequence version**. The FASTA header encodes it in the form sp|P04637|P53_HUMAN, and when a sequence changes for any reason (for example, a re-sequencing of the source genome corrects a residue), UniProt bumps the version and keeps the old sequence retrievable. This is why an accession alone is almost always sufficient in practice, but a version number makes your record airtight.

### Cross-references

Cross-references are the bridges from UniProt to the rest of biology, organized into groups:

- **Sequence databases** such as RefSeq and Ensembl.
- **3D structure databases** such as PDB and AlphaFold.
- **Protein-protein interaction databases** such as IntAct and STRING.
- **Protein family and domain databases** such as Pfam, PROSITE, and InterPro.
- **[Gene expression](/blog/guides/gene-expression) databases**.
- **Ontologies** such as the Gene Ontology.
- **Pathology and biotech resources** such as OMIM and DrugBank.
- **Chemistry databases** for enzymes with ligand data.
- **Genome annotation and other resources**, including organism-specific databases.

Two habits make this section powerful. First, use it to find the specialized resource for your next question. If UniProt's function summary is not enough for pathway work, the Reactome cross-reference takes you there. Second, notice that cross-references can be labeled as obsolete or removed. A PDB entry that was superseded or retracted may still appear with that status, which is informative rather than confusing once you know to look for it.

### Publications

UniProt lists the literature linked to the entry, with each citation tied to the specific annotations it supports. Where possible, curators tag publications with the topics they cover, such as function, characterization, or mutagenesis. The distinction between the **first author's own submissions** and independently published studies is preserved in the record, which helps you weigh self-reported claims.

### Entry information and history

At the bottom of the navigation you find the metadata that makes the entry citable:

- The **entry name** and all previous names it has carried.
- The **accession**, including any secondary accessions merged into this entry.
- The **entry version** and **sequence version**.
- The dates of creation and of the most recent modifications.
- The database the entry belongs to (UniProtKB/Swiss-Prot or UniProtKB/TrEMBL).

UniProt releases updates on a regular schedule, and entries change between releases. If you build a pipeline or write a methods section, capture the release and the entry version you used, so your results remain reproducible. UniProt supports programmatic retrieval that makes this straightforward, which we cover in the workflows section.

## How to Read Evidence and Quality Indicators

![Decision path asking whether the entry is reviewed, what its protein existence level is, and how strong the evidence behind a claim is.](/article-images/illustrations/uniprot-entry-page-tabs-explained-reviewed-triage.webp)

UniProt's defining design decision is that evidence attaches to claims, not just to entries. Three quality indicators deserve deliberate practice.

### Evidence tags

Hover over almost any annotation and a small tag appears, such as "Manual assertion according to rules" or "Manual assertion inferred by a curator" or "Automatic assertion according to sequence similarity." Click it and UniProt shows the supporting publication, the rule applied, or the source database. In the flat file view, the same information appears in evidence attributes on each line. The general rule for interpretation: assertions backed by named publications can be traced to data. Assertions "according to rules" or "by similarity" were derived computationally and may or may not have been checked against experiments for your specific protein.

### The annotation score

The annotation score summarizes how much is known about the entry, on a scale from 1 to 5. A score of 5 means the entry is richly annotated. Low scores accompany poorly studied proteins and most unreviewed records. The score is a coverage metric, not an accuracy metric. It tells you how much curation effort has been invested, not whether each annotation is correct, so use it to set expectations about depth.

### Protein existence

As described earlier, the PE level tells you whether the protein itself has been observed. A surprising number of genes predicted from genome assemblies encode proteins that have never been directly detected, and UniProt is honest about this. If you are about to design primers, antibodies, or structural models around a PE 4 entry, knowing that the sequence is purely predicted should shape your expectations.

## The Feature Viewer and Sequence Track

Along the top of the entry page, the **feature viewer** renders the sequence as a horizontal track with all positional annotations plotted along it. Categories such as processing, domains and sites, variants, and PTMs appear as colored tracks you can expand, collapse, and toggle. Clicking a feature scrolls to the corresponding sequence region and opens a detail panel with the evidence.

This view answers positional questions faster than reading any text section. Practical uses include:

- Seeing at a glance whether a disease variant falls inside a known domain.
- Checking whether your region of interest overlaps a transmembrane helix, which changes how you would express and purify it.
- Locating all known phosphorylation sites within a disordered tail.
- Seeing the boundaries of a low-complexity region before designing a construct.

For proteins with multiple isoforms, the viewer can display features on the isoform you select rather than only on the canonical sequence, which matters when the biology you care about lives in a splice-specific exon.

## A Worked Example: Reading the p53 Entry

Take the human p53 protein, accession P04637, entry name P53_HUMAN. The header tells you immediately that the entry is reviewed, the protein exists at the protein level (PE 1), and the annotation score is high, meaning deep curation. Length and mass are shown (393 amino acids, about 43.7 kDa).

Moving through the sections produces a coherent picture within a few minutes of reading:

- **Function** reports that p53 acts as a [transcription factor](/knowledge/molecular-biology/transcription-factor) central to the cellular response to DNA damage, with keywords such as "DNA-binding", "Transcription", "Apoptosis", and "Tumor suppressor", and GO terms linking it to processes such as regulation of cell cycle arrest.
- **Pathology & Biotech** lists the connection to Li-Fraumeni syndrome and a large catalog of missense, nonsense, and frameshift variants with references, including many hot-spot residues in the DNA-binding domain.
- **PTM / Processing** documents dozens of phosphorylation sites with the kinases reported to modify them, plus ubiquitination and acetylation events with their regulatory context.
- **Structure** links to numerous PDB entries, notably structures of the DNA-binding domain in complex with DNA, and to the AlphaFold full-length predicted model, which is useful for seeing the overall shape including the disordered terminal regions that no crystal structure captures. AFDB's current entry presentation pairs each predicted structure with an interactive viewer and domain summaries so that the predicted model can be interpreted alongside sequence annotations rather than in isolation [1].
- **Family & Domains** maps the transactivation domains, the DNA-binding domain, and the tetramerization domain with residue boundaries you can use directly for construct design.
- **Cross-references** hand you the keys to OMIM for clinical genetics, IntAct for interaction networks, and expression resources for tissue data.

The lesson generalizes to any protein. Read the header for confidence level, scan Function for a plain-language summary, then descend into the positional sections for the details your experiment actually requires.

## How to Use UniProt in Practical Workflows

Understanding the entry page is the foundation. The workflows below are the everyday uses where that understanding pays off.

### Workflow 1: From a gene name to the right protein

![Four step flow from searching a gene symbol through filtering by organism and reviewed status to recording the accession and sequence version.](/article-images/illustrations/uniprot-entry-page-tabs-explained-gene-to-protein.webp)

You start with a symbol like TP53 and need the human protein. Search UniProt with the gene symbol, filter results by organism, and restrict to reviewed entries when your downstream analysis depends on trustworthy annotation. The search panel supports fielded queries, for example gene names combined with taxonomy filters. Selecting the reviewed human entry gives you the canonical sequence and the full curated record.

When you need the corresponding proteins across species, the entry page provides links to orthologs within its similarity section, and the taxonomy filter in search lets you pull the reviewed entries for each organism in turn.

### Workflow 2: Getting a sequence for cloning or primer design

From any entry page, the Download button offers the FASTA format sequence. FASTA is the universal flat format, a header line beginning with ">" followed by the sequence in single-letter amino acid code. Choose the canonical sequence or a specific isoform, copy it into your cloning software or oligo design tool, and record the accession and sequence version in your notebook. This two-minute habit is what makes a construct traceable three years later.

### Workflow 3: Checking variants against function

Suppose a genome-sequencing report gives you a p53 variant, R175H. On the entry page, use the variants feature in the viewer or the Pathology & Biotech section to find residue 175, read what mutations there have been shown to do, and follow the citations. The cross-references to clinical variant databases extend the picture. This workflow, residue in hand to literature in two clicks, is one of the strongest arguments for making UniProt a first stop.

### Workflow 4: Mapping identifiers across databases

Real datasets arrive with mixed identifiers. UniProt's **ID mapping** tool converts a list of gene names, RefSeq accessions, Ensembl transcript IDs, or other identifiers into UniProt accessions. Paste your list, choose the source and target identifier types, and download the mapped table. For bulk work, the same capability is available through the REST API, so a script can map thousands of identifiers without manual copy-paste.

### Workflow 5: Programmatic retrieval

Every entry is available without the website through the REST API. A single command returns the FASTA sequence for an accession:

```
curl https://rest.uniprot.org/uniprotkb/P04637.fasta
```

Requests can also return tabular fields, such as protein names, organism, and length, and full text formats for complete records. Queries can filter by reviewed status, organism, annotation type, and more, which is how you build reproducible pipelines on top of curated data. The API and its documentation are maintained by UniProt as part of the resource itself, so the exact parameters and formats are documented on the UniProt help pages. When a project requires structure context as well as sequence, the AlphaFold database exposes its models through its own website, FTP, cloud services, and APIs, and its predictions are aligned to UniProt releases so that sequences and structures can be joined by accession without manual reconciliation [1].

### Workflow 6: Preparing for structural analysis

Before committing an experiment or a modeling project, the Structure section lets you check whether an experimental structure already exists for your protein or a close homolog, what coverage it has, and what the AlphaFold prediction suggests about domain organization. Combining the domain boundaries from Family & Domains with the coverage maps in Structure tells you whether the region you care about is structurally characterized, structurally predictable, or genuinely unknown, which is exactly the decision you need before choosing between homology modeling, structure prediction, or experimental determination.

### Why curation quality matters for these workflows

The value of every workflow above depends on the underlying data being findable, consistent, and traceable to origin. That is a general problem in biomedical informatics. Public resources differ in formats and query mechanisms, and manually collecting equivalent facts about a protein from many separate sites produces information of variable quality that is hard to compare and easy to mis-attribute. Curated hubs like UniProt, and structure resources that organize their records around the same protein identifiers, exist precisely to reduce that friction, and the recent redesign of large-scale structure resources around summary and domain tabs with annotations integrated into an interactive viewer is a direct response to the same usability challenge [1]. The entry page is not just convenient. It is the quality-control layer between raw sequence data and your interpretation.

## Common Mistakes and Limitations

The mistakes below account for most of the errors seen in student projects, manuscripts under review, and pipelines that mysteriously produce inconsistent results.

1. **Citing the entry name instead of the accession.** Entry names can change. If your paper cites P53_HUMAN rather than P04637, and UniProt later replaces the name, your reference degrades silently. Accessions are the currency of reproducibility.

2. **Treating unreviewed entries as curated.** A TrEMBL record's function section may be entirely machine-generated. It is not wrong by design, but it is not verified. For claims you intend to publish, start from reviewed entries or verify the automatic annotations yourself.

3. **Ignoring evidence tags.** "By similarity" and "potential" qualifiers exist for a reason. Students frequently copy an annotation into a report, drop the qualifier, and thereby convert a prediction into an asserted fact. Keep the qualifier or verify the primary source.

4. **Using the wrong isoform.** If your experiment targets an isoform expressed in your tissue and you clone the canonical sequence, your construct may lack the exon you needed. Read the isoform descriptions and check expression resources when splicing matters.

5. **Confusing experimental and predicted structures.** A PDB model is a measurement of some state of some construct, often truncated. An AlphaFold model is a prediction with per-residue confidence. Neither should be described with the other's language in a paper.

6. **Ignoring entry and sequence versions.** Between releases, entries gain annotations, merge duplicates, and occasionally correct sequences. Pipelines that fetch "whatever is current" produce irreproducible outputs. Pin the release and record the version.

7. **Over-reading annotation scores.** A high score means rich annotation, not guaranteed accuracy. A low score means limited curation, not a boring protein. Both require reading the evidence, not just the number.

8. **Assuming absence of annotation means absence of biology.** Poorly studied proteins exist in enormous numbers. The blank sections of an unreviewed entry are gaps in knowledge, not evidence of non-function.

The broader limitation is inherent to all databases of this kind. UniProt reflects the current state of experimental literature and computational prediction, both of which evolve. Annotations are revised, structures are superseded, and cross-references change. Individual research decisions, especially clinical or experimental ones, deserve verification against the primary literature that the entry page conveniently links to but does not replace.

## Quick Review

- A UniProt entry page is the curated record for one protein sequence, with evidence attached to individual claims rather than the entry as a whole.
- The accession is the permanent identifier, the entry name is a readable label that can change, and the entry version plus sequence version make records reproducible.
- Reviewed (Swiss-Prot) entries are manually curated, unreviewed (TrEMBL) entries are automatically annotated, and the gold or silver badge tells you which you are reading.
- Protein existence levels 1 through 5 and the annotation score set your expectations before you read the details.
- Positional sections (PTM/Processing, Pathology & Biotech, Family & Domains, Structure) answer residue-level questions, and the feature viewer makes those answers visual.
- Evidence tags, especially the difference between experimental and electronic annotation, determine how strongly you can state any conclusion.
- Cross-references are the bridges to OMIM, PDB, IntAct, Reactome, and dozens of specialized resources, and the REST API turns the whole knowledgebase into a scriptable endpoint.

## Frequently Asked Questions

### What does "reviewed" mean on a UniProt entry page?

It means the entry belongs to the Swiss-Prot component of UniProtKB and has been manually annotated by professional curators who verified the information against the scientific literature. Unreviewed entries belong to TrEMBL and carry automatic, rule-based annotation instead.

### What is the difference between the accession and the entry name?

The accession, such as P04637, is a stable permanent identifier that never changes and never gets reused. The entry name, such as P53_HUMAN, is a human-readable label that UniProt may replace when nomenclature improves. Cite the accession for reproducibility.

### How current are UniProt entries and how do I cite a specific version?

UniProt updates entries on a regular release schedule, and each entry carries an entry version and a sequence version that change when content changes. Record the release you used and the entry version, and for programmatic work pin your queries to a specific release so results remain reproducible.

### Can I trust annotations marked "by similarity" or with electronic evidence?

Treat them as starting points rather than established facts. They were transferred computationally from related proteins or derived by rules rather than demonstrated experimentally for this specific protein. For published claims, follow the evidence tag to its source or search for direct experimental data.

### How do I download a sequence in FASTA format?

Open the entry page, use the Download button, and choose FASTA for the canonical sequence or a selected isoform. For scripts, request the same format directly from the REST API with the accession in the URL. Record the accession and sequence version alongside the sequence.

### What is the difference between UniProtKB, UniRef, and UniParc?

UniProtKB holds the annotated knowledgebase entries you read on entry pages, both reviewed and unreviewed. UniRef clusters sequences at 100%, 90%, and 50% identity to reduce redundancy for fast searches. UniParc is a comprehensive archive of all unique sequences with identifiers but no biological annotation.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "What does \"reviewed\" mean on a UniProt entry page?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "It means the entry belongs to the Swiss-Prot component of UniProtKB and has been manually annotated by professional curators who verified the information against the scientific literature. Unreviewed entries belong to TrEMBL and carry automatic, rule-based annotation instead."
      }
    },
    {
      "@type": "Question",
      "name": "What is the difference between the accession and the entry name?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "The accession, such as P04637, is a stable permanent identifier that never changes and never gets reused. The entry name, such as P53_HUMAN, is a human-readable label that UniProt may replace when nomenclature improves. Cite the accession for reproducibility."
      }
    },
    {
      "@type": "Question",
      "name": "How current are UniProt entries and how do I cite a specific version?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "UniProt updates entries on a regular release schedule, and each entry carries an entry version and a sequence version that change when content changes. Record the release you used and the entry version, and for programmatic work pin your queries to a specific release so results remain reproducible."
      }
    },
    {
      "@type": "Question",
      "name": "Can I trust annotations marked \"by similarity\" or with electronic evidence?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Treat them as starting points rather than established facts. They were transferred computationally from related proteins or derived by rules rather than demonstrated experimentally for this specific protein. For published claims, follow the evidence tag to its source or search for direct experimental data."
      }
    },
    {
      "@type": "Question",
      "name": "How do I download a sequence in FASTA format?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Open the entry page, use the Download button, and choose FASTA for the canonical sequence or a selected isoform. For scripts, request the same format directly from the REST API with the accession in the URL. Record the accession and sequence version alongside the sequence."
      }
    },
    {
      "@type": "Question",
      "name": "What is the difference between UniProtKB, UniRef, and UniParc?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "UniProtKB holds the annotated knowledgebase entries you read on entry pages, both reviewed and unreviewed. UniRef clusters sequences at 100%, 90%, and 50% identity to reduce redundancy for fast searches. UniParc is a comprehensive archive of all unique sequences with identifiers but no biological annotation."
      }
    }
  ]
}
</script>

## Sources

1. [AlphaFold Protein Structure Database 2025: a redesigned interface and updated structural coverage.](https://pubmed.ncbi.nlm.nih.gov/41273079/)