UniProt through GeneFoundry

Router invocation: uniprot_get_server_capabilities()

Return the uniprot-link discovery surface. detail='summary' (default) is light: identity/build/release, the tool list WITH call signatures, accepted argument aliases, response modes, recommended workflows, error taxonomy, and limits -- enough to call any tool without guessing an argument name. detail='full' adds the heavy reference blocks (21 named graphs with triple counts, the full SPARQL prefix map, full latency bands, feature-type and cross-reference vocabularies). Call this first in a cold session, or read uniprot://tools (signatures only) or uniprot://capabilities (full). Signature: get_server_capabilities(detail=).

Router invocation: uniprot_search_sparql_query()

Execute a bounded SPARQL SELECT/ASK query against the UniProt endpoint (bounded SELECT / ASK only). SERVICE federation and graph-returning CONSTRUCT/DESCRIBE forms are rejected. SELECT results come back as columns+rows JSON; ASK as a boolean. A LIMIT is auto-injected into unbounded SELECTs (see `_meta`/`truncated`). This is the escape hatch for anything the typed tools do not cover -- seed queries from search_example_queries. Use uniprot://prefixes for the standard PREFIX block. Broad, unanchored queries can take 10-60 s; bound lookups (anchored on an accession/gene/taxon) return in <2 s. Signature: search_sparql_query(query, result_format=, limit=, timeout_seconds=).

Router invocation: uniprot_search_example_queries()

Search UniProt's 126 curated, executable SPARQL example queries by free text over their descriptions and keyword tags (e.g. 'disease', '3D structure', 'cross-reference', 'taxonomy'). Returns example ids, descriptions, tags, and query types. Fetch the full query text with get_example_query, then run it via search_sparql_query. The best way to learn how to query UniProt. Signature: search_example_queries(text=, limit=).

Router invocation: uniprot_get_example_query()

Fetch one curated example's full SPARQL text, description, keyword tags, and any federated endpoints it joins. Pass an example_id (full IRI) from search_example_queries. `_meta.next_commands` offers to run it directly via search_sparql_query. Signature: get_example_query(example_id).

Router invocation: uniprot_find_proteins()

Search UniProtKB for the entries of a GENE and return matching entries (accession, mnemonic, recommended name, reviewed flag, organism). gene_symbol is REQUIRED; refine the hits with organism_taxon (a positive NCBI taxon id or an exact common/scientific organism name), reviewed, name_contains (matched per word, in any order, case-insensitive), or the structured filters mnemonic / ec_number / keyword. Reviewed (Swiss-Prot) hits are ranked first. UniProt SPARQL has no general full-text index; for EC-only, keyword-only, or free-text search WITHOUT a gene, use search_example_queries or search_sparql_query. Pair with get_protein for full detail. Results are ordered reviewed-first, then by mnemonic, then accession (stable across pages). Cold search can take several seconds; an identical repeat is cached (~0 ms). If you already know the accession, call get_protein directly -- it is far faster. Signature: find_proteins(gene_symbol, organism_taxon=, reviewed=, keyword=, ec_number=, mnemonic=, name_contains=, limit=, offset=).

Router invocation: uniprot_find_proteins_batch()

Resolve SEVERAL gene symbols to UniProtKB entries in ONE call, running the lookups concurrently -- so N genes cost about one cold round-trip instead of N sequential ones. Use this for multi-gene tasks (e.g. 'get domains for PNKP and NAA10'). Returns by_gene (gene -> accessions, reviewed-first), a flat proteins list tagged with matched_gene, resolved_genes, and unresolved_genes (a symbol that matched nothing is disclosed, never silently dropped). Optionally scope by organism_taxon (a positive NCBI taxon id or an exact common/scientific organism name) and reviewed. next_commands fan out to get_protein on each resolved gene's top hit. For a single gene use find_proteins. Signature: find_proteins_batch(gene_symbols, organism_taxon=, reviewed=, limit_per_gene=).

Router invocation: uniprot_get_protein()

Return the core summary for a single UniProtKB entry by accession: mnemonic, reviewed flag, recommended/short name, gene(s), organism + taxon, protein existence, sequence length and mass, a function summary, and creation/modification dates, plus has_variants/has_diseases/has_structure presence flags that drive content-aware next_commands. An obsolete/demerged accession returns a flagged obsolete record (obsolete:true + replaced_by). response_mode (default compact) controls verbosity; standard/full add the created/modified dates. Signature: get_protein(accession, response_mode=).

Router invocation: uniprot_get_protein_sequence()

Return the amino-acid sequence(s) for an entry: the canonical isoform (length, mass, sequence) plus any additional (non-canonical) isoforms. Pass a canonical accession for all isoforms, or an isoform accession (e.g. P05067-2) to get THAT isoform's specific sequence and mass. response_mode controls verbosity: minimal=metadata only; compact (default)=length/mass + a first/last-30-residue sequence_preview (sequence_truncated:true) — cheap for large proteins; standard/full return the complete sequence string. Set canonical_only=true to return only the canonical isoform (skip the additional-isoform list). Signature: get_protein_sequence(accession, response_mode=, canonical_only=).

Router invocation: uniprot_get_protein_features()

Return sequence features with begin/end coordinates (FALDO) for an entry: domains, regions, transmembrane segments, binding/active sites, PTMs, signal peptides, secondary structure, mutagenesis sites, and more. feature_types=['domain'] returns positional domain extents; each returned `type` round-trips to the filter vocabulary. Filter keys come from capabilities (feature_types); a zero-match filter echoes the accepted keys as a filter_hint. Secondary-structure features (helix/strand/turn) are hidden by default and disclosed under excluded_secondary_structure; set include_secondary_structure=true (or name them in feature_types) to return them. DNA-binding extents are not available from UniProt RDF, so dna_binding is deliberately not a filter. response_mode=compact/minimal drops repeated fenced descriptions; standard (default) preserves existing full records. Signature: get_protein_features(accession, feature_types=, limit=, include_secondary_structure=, response_mode=).

Router invocation: uniprot_get_protein_variants()

Return natural-variant annotations for an entry: position, wild-type residue, amino-acid substitution, an HGVS-style `notation` (e.g. `L176F`) for simple substitutions, `variant_type` (substitution|other), free-text description, structured linked `diseases`, and `dbsnp` rsIDs. Set disease_associated_only=true to keep only disease-linked variants. Set position_start and/or position_end to keep annotations that overlap that residue interval (inclusive). response_mode=compact/minimal drops the repeated free-text provenance; standard (default) preserves the existing full records. Signature: get_protein_variants(accession, limit=, disease_associated_only=, response_mode=, position_start=, position_end=).

Router invocation: uniprot_get_protein_diseases()

Return disease annotations associated with an entry: disease name, UniProt disease id, mnemonic, MIM id, the clinical `definition` (the disease vocabulary's own description), and `involvement` (the entry-specific note). Pairs with get_protein_variants for variant-level disease evidence. response_mode=compact/minimal drops the free-text definition/involvement; standard (default) preserves them. Signature: get_protein_diseases(accession, response_mode=).

Router invocation: uniprot_get_protein_cross_references()

Return database cross-references for an entry, grouped by database (PDB, AlphaFoldDB, Ensembl, RefSeq, Reactome, STRING, InterPro, ...). Optionally restrict to specific databases (case-sensitive); any requested name that matched nothing is echoed under unmatched_databases with a did-you-mean, so a typo never reads as 'no data'. response_mode (default compact) returns short ids; full restores raw IRIs. Returns every cross-reference database; use resolve_identifiers for a focused primary-id mapping. Signature: get_protein_cross_references(accession, databases=, response_mode=).

Router invocation: uniprot_get_protein_go_terms()

Return Gene Ontology annotations for an entry, grouped by aspect (biological_process / molecular_function / cellular_component) where available, each with GO id, label, and (when annotated) ECO `evidence` ids plus mapped GO `evidence_codes` (IDA/IEA/IMP/...) for citation. Always returns `count` and `count_by_aspect`; pass `aspect` to scope to one ontology and `limit` to cap a large set (token economy). Signature: get_protein_go_terms(accession, aspect=, limit=).

Router invocation: uniprot_resolve_identifiers()

Resolve a UniProtKB accession to its PRIMARY external identifiers: the genomic/structural/family core (PDB, AlphaFoldDB, Ensembl, RefSeq, GeneID, HGNC, KEGG, OrthoDB, Pfam, InterPro) by default. Optionally restrict to specific databases. Returns ids grouped by database plus the databases that matched and per-database counts. response_mode (default compact) returns short ids; full restores raw IRIs. For the exhaustive cross-reference set (incl. drug/disease databases like DrugBank/ChEMBL/OpenTargets) use get_protein_cross_references instead. Signature: resolve_identifiers(accession, databases=, response_mode=).

Router invocation: uniprot_get_taxon()

Resolve an organism in the UniProt taxonomy. Pass a numeric NCBI taxon id (e.g. 9606) for full detail (scientific/common name, rank, the DIRECT parent, and an optional ordered lineage from species up to root), or a scientific/common name to get candidate taxon ids. Use the resolved taxon id with find_proteins(gene_symbol=..., organism_taxon=...). Name matches are ranked best-first (an exact scientific/common-name hit leads, tagged match_quality:'exact'), so matches[0] and next_commands point at the right organism. Numeric-id and common-organism-name lookups are fast (~0 ms for common names); an uncommon name triggers a multi-second taxonomy scan. Signature: get_taxon(taxon, include_lineage=).