UniProt Reference

Swiss-Prot (reviewed:true) contains approximately 570,000 protein entries that have been manually curated and reviewed by expert biologists, with experimentally verified annotations and literature references. TrEMBL (reviewed:false) contains approximately 250 million entries that are automatically annotated by computational pipelines without manual review. Swiss-Prot entries are considered high-confidence and are preferred for research, while TrEMBL provides broader coverage including computationally predicted proteins.

Use the REST API at https://rest.uniprot.org/uniprotkb/search with query parameters. For example: ?query=(gene:BRCA1)+AND+(organism_id:9606)&format=json&size=10 finds human BRCA1 entries in JSON format. You can filter by fields (accession, protein_name, gene_names, organism_name), specify format (json, tsv, fasta, xml), and use pagination (size, cursor). The &fields= parameter returns only selected columns, reducing bandwidth for large queries.

PE levels indicate the type of evidence supporting a protein's existence: PE 1 = experimental evidence at protein level (mass spectrometry, X-ray, etc.), PE 2 = evidence at transcript level (mRNA detected but protein not directly confirmed), PE 3 = inferred by homology from a characterized protein in another species, PE 4 = predicted by gene prediction algorithms, PE 5 = uncertain/dubious. For high-confidence analyses, filter for PE 1 entries.

Use the ID mapping API: POST to https://rest.uniprot.org/idmapping/run with parameters specifying the source database (from=GeneCards, Ensembl, PDB, RefSeq, etc.) and target (to=UniProtKB). The API returns a job ID; poll https://rest.uniprot.org/idmapping/results/{jobId} for results. This handles batch conversions and supports dozens of external databases. The REST API response includes mapped accessions with metadata.

A UniProt FASTA header follows this structure: >db|accession|entry_name Description OS=Organism OX=TaxonomyID GN=GeneName PE=ProteinExistence SV=SequenceVersion. For example: >sp|P04637|P53_HUMAN Cellular tumor antigen p53 OS=Homo sapiens OX=9606 GN=TP53 PE=1 SV=4. The "sp" prefix indicates Swiss-Prot (manually reviewed), while "tr" indicates TrEMBL (auto-annotated). This header format is standardized and parseable by most bioinformatics tools.

Feature types annotate specific regions or positions in the protein sequence. Key types include: DOMAIN (functional domain boundaries), BINDING (ligand/substrate binding sites), ACT_SITE (catalytic active site residues), MOD_RES (post-translational modifications like phosphoserine), VARIANT (natural sequence variants with dbSNP IDs and clinical significance from ClinVar), MUTAGEN (experimentally introduced mutations and their effects), CHAIN (mature protein after signal peptide cleavage), and TRANSMEM (transmembrane helix regions).

UniRef provides pre-computed sequence clusters at three identity thresholds: UniRef100 (identical sequences merged), UniRef90 (90%+ identity), and UniRef50 (50%+ identity). Use UniRef90 or UniRef50 to reduce redundancy in sequence databases for faster BLAST searches, to create non-redundant training sets for machine learning, or to reduce computational costs in large-scale analyses. Each cluster has a representative sequence that is typically the best-annotated Swiss-Prot entry.

Use the requests library to call the REST API: import requests; resp = requests.get("https://rest.uniprot.org/uniprotkb/P04637.json"); data = resp.json(). The JSON response contains structured data including proteinDescription, gene, organism, comments (function, subcellular location), features (domains, variants), and cross-references. For bulk operations, use the stream endpoint with format=tsv or format=fasta and write directly to files. The requests library handles pagination via the Link header for search results.