LSMA

LONDON SCHOOL OF MULTIOMICS & AI

Free course

Free course

0 of 73 lessons

Lesson 4 of 7

Module 2 · Biological data and databases

Reading a gene record and a protein record

Reading a gene record and a protein record

Reading a gene record and a protein record

Gene and protein pages look crowded, but every one follows the same plan. Walk through a gene record and a UniProt entry for BRCA1, and learn to place a variant on the protein and judge how far to trust what you read.

⏱ 13 min readModule 2 · Biological data and databasesFree
Lesson at a glance4 steps · about 13 min
  1. Records share a skeleton
  2. A gene record, section by section
  3. A protein record in UniProt
  4. Can you trust it? Review status and evidence

Records share a skeleton

Identifiers and cross-references: following one gene across databases showed how to find the right record. This lesson shows how to read one. Gene and protein pages look crowded at first, but each database uses the same sections for every gene, in the same order. Learn the layout once and you can read any record in a few minutes.

In Identifiers and cross-references you learned to tell a gene label from a transcript or protein label; here those labels become the headings of real pages. The skeleton matters because you will rarely read a record from top to bottom. Usually you come with one question, such as "which transcript?" or "what does residue 61 do?", and knowing where each answer sits lets you go straight to it. Websites are redesigned from time to time, so buttons and colours may not match what you see here, but the sections and what they hold stay the same.

We will use BRCA1 throughout. It is the gene in which inherited harmful variants raise the risk of breast and ovarian cancer, and its protein helps repair broken DNA. Its main identifiers are HGNC:1100, NCBI Gene ID 672, Ensembl ENSG00000012048 and UniProt P38398, and it lies on chromosome 17.

A well-studied gene is the best place to learn, because every section is filled in and the evidence is strong. Records for less-studied genes have the same skeleton but more empty sections, and learning what a full record looks like makes it easier to see what is missing from a sparse one.

A gene record, section by section

NCBI Gene and Ensembl lay their gene pages out differently, but they cover the same ground. The table lists the sections you will meet, in roughly the order they appear.

SectionWhat it tells youFor BRCA1
SummaryApproved symbol, full name, aliases, gene type, a short description"BRCA1 DNA repair associated"; aliases include RNF53 and FANCS; protein coding
LocationChromosome, band, strand, and which genome build the positions useChromosome 17, long arm; positions given for GRCh38
TranscriptsEvery known transcript, its biotype and its protein, if anyMany transcripts, several of them coding
Reference transcriptThe one transcript chosen as the default, flagged MANE SelectOne RefSeq NM_ and Ensembl ENST pair
ExonsThe exons of a chosen transcript, with their sizesOne unusually large exon, traditionally called exon 11
OrthologuesThe matching gene in other speciesBrca1 in mouse, and many more
LinksCross-references to proteins, variants, diseases, expression and papersUniProt, ClinVar, OMIM, PubMed and others

How to read this table

Each row is one block of the page. The middle column says what that block answers; the right column shows what you would see for BRCA1. Work down the rows in order: first confirm you have the right gene, then where it is, then which of its transcripts matters, then how it relates to other species and other databases.

  • The Summary is where you check you have the right gene. Compare the symbol and species with what you expected, and read the alias list: older papers may call BRCA1 by one of them. The gene type, or biotype, says what kind of gene it is, such as protein coding, non-coding RNA or pseudogene (a broken copy of a gene that no longer makes a working protein).
  • Location gives the chromosome and the cytogenetic band, the region seen as a stripe on a stained chromosome under the microscope. The exact positions belong to one genome build (Identifiers and cross-references: following one gene across databases), and the page states which.
  • The transcript table lists every version of the RNA the gene is known to make. In Ensembl each row has an ENST identifier, a length, a biotype and, if it codes for a protein, an ENSP identifier. Some transcripts are fragments or are kept for further study; not all are equally well supported.
  • MANE Select (Matched Annotation from NCBI and EMBL-EBI) is a single transcript per protein-coding gene, agreed by both teams. Its RefSeq and Ensembl versions match base for base and match the GRCh38 genome exactly. Clinical laboratories increasingly describe variants against it, so it is the transcript to choose unless you have a reason not to.
  • Exons are the stretches kept in the mature RNA; introns are the stretches spliced out between them. Exon numbers can differ between databases and older papers. BRCA1 is a known case: its traditional numbering skips exon 4, a leftover from early work, so always check which numbering a source uses.
  • An orthologue is the same gene in another species, descended from one gene in a common ancestor. Orthologues tell you whether a mouse model is relevant and which parts of the protein have been kept through evolution. A paralogue, by contrast, is a related gene in the same species, made by an old duplication.

One detail in the Location section often confuses new readers: the strand. Genome positions always count along the forward (+) strand, from the top of the chromosome down. BRCA1 lies on the reverse (−) strand, so it is read in the direction of falling positions: its first exon sits at the higher coordinate and its last exon at the lower one. A tool or a person who forgets this will read the exons back to front. The summary text is also worth a second look: on NCBI Gene it is often written by RefSeq curators and cites the papers it draws on, so it is a reasonable first account of what the gene does, but not a replacement for those papers.

NCBI Gene
  • One long page, sections stacked
  • Strong on papers: links to PubMed and short curated notes
  • Links to RefSeq, ClinVar and dbSNP
Ensembl gene page
  • Tabs down the left: summary, transcripts, comparative genomics, variants
  • Strong on transcripts, biotypes and gene trees
  • Genome browser view of the region

MANE is described by Morales and colleagues (2022, Nature). Its first full release, in 2022, covered about 19,000 human protein-coding genes with one MANE Select transcript each. A small extra set, MANE Plus Clinical, adds a second transcript for genes where clinically important variants fall outside the Select transcript.

Two cautions. MANE versions are released over time, so a report may name an older transcript version than the one you see today; compare the version numbers. And MANE Select is a default, not a statement that other transcripts are unimportant. If a report uses a different transcript, its c. and p. positions may legitimately differ from the MANE ones.

A protein record in UniProt

A gene record is about the stretch of DNA. A protein record is about what the protein does, collected from experiments and papers. UniProt is the place to read it. The figure shows the layout of an entry as a schematic.

Reading a gene record and a protein record · LSMA free course A schematic mock-up, not a screenshot. A header shows accession P38398, entry name BRCA1_HUMAN, a Reviewed badge, the protein name and the species. A column on the left lists the sections in order: Function, Names and taxonomy, Subcellular location, Disease and variants, PTM and processing, Structure, Family and domains, Sequence and isoforms. The main panel shows the feature viewer: a bar for the 1,863-residue protein with a RING domain near the start, two BRCT domains at the end and a long middle region with few folded domains. A marked variant sits inside the RING domain. Below, a structure track shows short experimental structures covering the RING and BRCT domains, and one predicted model covering the whole chain. P38398 · BRCA1_HUMAN Reviewed Breast cancer type 1 susceptibility protein · Homo sapiens · 1,863 residues SECTIONS Function Names & taxonomy Subcellular location Disease & variants PTM / processing Structure Family & domains Sequence & isoforms FEATURE VIEWER Domains RING BRCT × 2 long middle region, few folded domains 1 1,863 Variants a variant of interest Structures experimental (PDB) predicted model of the whole chain (AlphaFold) Evidence tags on every statement: experimental, by similarity, or automatic
The parts of a UniProt entry you will use most. The section names are close to those on the real site; the drawing, spacing and domain sizes are simplified and positions are approximate. Schematic, not a screenshot.

How to read this figure

Start with the header. It gives the accession (the identifier to cite), the entry name, whether the entry is reviewed, the protein name, the species and the length in residues, meaning amino acids. Then use the section list on the left as a table of contents. The main panel is the feature viewer, which draws the protein as a bar from its first residue on the left to its last on the right and stacks tracks of information along it.

  • Function is a written summary of what the protein does, backed by references. For BRCA1 it describes a role in repairing breaks in both DNA strands by homologous recombination, a repair method that copies the missing information from the matching chromosome, and an enzyme activity that tags other proteins with ubiquitin, working as a pair with a partner protein, BARD1.
  • Names and taxonomy list the recommended protein name, alternative names and the gene that encodes it. BRCA1 is also known as E3 ubiquitin-protein ligase BRCA1.
  • Subcellular location says where in the cell the protein is found. BRCA1 works mainly in the nucleus.
  • Features are positions along the chain, given as residue numbers. Domains are parts that fold on their own and often have a job of their own. Active sites are the residues that do the chemistry of an enzyme. Modified residues carry a chemical tag added after the protein is made, such as a phosphate; this is post-translational modification, or PTM. BRCA1 is phosphorylated at many sites in response to DNA damage.
  • Disease and variants lists natural variants, the change at each position and any disease link, with links to variant databases such as ClinVar.
  • Structure lists experimental 3D structures in the Protein Data Bank (PDB), with the residues each one covers, and a link to the predicted model from the AlphaFold database.
  • Sequence and isoforms shows the canonical sequence, the one all positions in the entry refer to, and any other isoforms.

The canonical sequence deserves care. UniProt chooses one isoform per entry as canonical, usually the one that is most common or best described, and every feature position in the entry counts along it. For most human genes this protein matches the one made from the MANE Select transcript, but not for every gene. When positions from a clinical report and from UniProt disagree by a few residues, a different isoform is the usual cause, and comparing the two lengths, as in the worked example below, is the quickest check.

Can you trust it? Review status and evidence

UniProt has two parts. Swiss-Prot entries are reviewed: a curator has read the literature and written or checked each statement. TrEMBL entries are unreviewed: they are made automatically from sequences, and their annotation is predicted. Swiss-Prot is much smaller, but far more reliable. BRCA1's entry is reviewed.

Within an entry, each statement carries an evidence tag. It says whether the statement comes from an experiment reported in a paper, was inferred by similarity to a related protein, or was added by an automatic rule. Each entry also has an annotation score from 1 to 5, a rough measure of how much is known, and a protein existence level, which says whether the protein itself has been detected or is only predicted from a transcript or a gene. Treat a claim inferred by similarity as a good guess, not a finding.

Why does this matter in practice? Automatic annotation spreads: if one protein is misannotated, proteins that resemble it can inherit the same claim, and the error then looks well supported because many entries repeat it. Evidence tags let you find where a statement really came from. A good routine for any claim you plan to rely on is to click its evidence tag, see whether it points to a paper, and read at least that paper's abstract. Even a reviewed entry reflects the literature at the time of its last update. UniProt releases new versions several times a year, so for a fast-moving protein, check recent papers too and note the release you used.

The annotation score needs the same care. A score of 5 means the entry is rich in information, not that every statement in it is correct, and a score of 1 for a new protein often just means that few papers exist yet. Use the score to judge how much there is to read, and the evidence tags to judge how far to trust each statement.

Worked example · a variant, placed on the protein

A report names the BRCA1 change p.Cys61Gly, a well-studied variant. In HGVS protein notation, p. means protein, so this reads "the cysteine at position 61 is replaced by glycine". Use the UniProt entry to answer: where is it, in which domain, and is there a structure?

1. Check the numbering: report's protein length = 1,863 = UniProt canonical length ✓
2. Features: RING-type zinc finger near the start, roughly residues 20 to 65
3. Position 61 falls inside it → in the RING domain
4. Structure: experimental structure of the RING region (with BARD1) ✓; whole-chain predicted model ✓

Step by step. First make sure the report and UniProt count along the same sequence. The report should name its transcript, ideally MANE Select; its protein length should match UniProt's canonical sequence. If the lengths differ, the report may be using another isoform and every position could be off. Next open the Family and domains section, or the feature viewer, and find the feature whose start and end include residue 61. For BRCA1 that is the RING domain, a small fold near the start that is held together by zinc ions gripped by cysteines and histidines. (Read the exact start and end from the live entry; the range above is approximate.) Then open the Structure section and look at which residues each PDB entry covers. Experimental structures exist for the RING region, bound to BARD1, and separately for the BRCT pair at the far end. No experimental structure covers the whole protein, but the predicted model does.

What the answer means. The variant removes a cysteine inside a zinc-binding fold, and the RING domain is what lets BRCA1 pair with BARD1 and tag other proteins. Losing one of the zinc-gripping residues would be expected to disturb that fold, which fits with this change being studied as harmful. That is a reasoned hypothesis from the record, not a diagnosis: clinical classification uses many lines of evidence and lives in resources such as ClinVar. When you open the predicted model, check its confidence colouring. The RING and BRCT domains are usually predicted well, while the long middle region of BRCA1 is mostly unstructured and gets low confidence; that low-confidence part should not be read as a real shape.

Go deeper: the five levels of protein existence

Every UniProt entry carries a protein existence level from 1 to 5. Level 1 means the protein itself has been detected, for example by mass spectrometry or in a structure. Level 2 means only its RNA has been seen. Level 3 means it is inferred because related proteins exist in other species. Level 4 means it is predicted from the genome alone, and level 5 means even the gene is doubtful.

The evidence tags on each statement use a shared vocabulary, the Evidence and Conclusion Ontology (ECO), so that a computer can filter by them. A statement backed by a curator reading an experiment in a paper carries a code such as ECO:0000269, with the paper's PubMed identifier beside it. Clicking the tag on the website shows both. For a strong claim, look for that combination: a reviewed entry, existence level 1 and an experimental tag.

Key point

A gene record tells you which transcript to trust; a protein record tells you what each position does and how good the evidence is. Read them together, and check that both count along the same sequence.

A second example: an unreviewed entry for a little-studied protein lists a kinase domain "by similarity" and no experimental structure. That is a useful starting point for an experiment, not proof that the protein is a working kinase. Look for a reviewed entry, a paper, or at least a confident predicted model before relying on it.

Further reading · Module 2MANE project: how the agreed reference transcripts are chosen. · UniProt help: every section of an entry explained. · UniProt evidence: what the evidence tags mean. · PDB-101: an introduction to 3D structures. · All reading for this course

Check your understanding

1. A gene has twelve transcripts listed. You need one to describe a variant for a clinical report. Which is the sensible default?

2. A report places a variant at residue 412, but its protein is 20 residues shorter than UniProt's canonical sequence. What should you do first?

3. An unreviewed UniProt entry lists a domain with the evidence tag "by similarity". How should you treat it?

Up next

Lesson 5

Public data and reading a paper

Ask BioTutor

Create a free website with Framer, the website builder loved by startups, designers and agencies.