aurekaresearch/opendde-harness/uniprot-lookup

UniProt lookup

Fixed, keyless endpoints. Fetch them with web_fetch; JSON, TSV and FASTA come back as is. Always ask for fields -- a whole entry is 60KB+ and gets truncated, a field selection is a few hundred bytes.

Find the accession

https://rest.uniprot.org/uniprotkb/search?query=gene:PDCD1+AND+organism_id:9606+AND+reviewed:true&fields=accession,protein_name,length&format=tsv&size=5

Query syntax: gene:, protein_name:, organism_id: (9606 human, 10090 mouse), accession:, reviewed:true (Swiss-Prot only). Combine with AND.

Read an entry

https://rest.uniprot.org/uniprotkb/Q15116?fields=accession,protein_name,length,sequence,ft_domain,ft_signal,ft_chain,ft_topo_dom,ft_transmem,ft_carbohyd,xref_pdb&format=json

Field names: sequence, ft_domain (domains with positions), ft_signal, ft_chain (mature chain), ft_topo_dom (the regions outside the membrane: extracellular / cytoplasmic), ft_transmem (the membrane-spanning segments), ft_carbohyd (glycosylation sites), ft_disulfid, xref_pdb (PDB entries with chains and ranges), cc_function. Positions are 1-based on the full precursor sequence, signal peptide included.

FASTA only:

https://rest.uniprot.org/uniprotkb/Q15116.fasta

Cutting an extracellular domain

Read ft_signal and ft_topo_dom from the entry, then slice the sequence from ft_topo_dom (Extracellular) start to end; do not guess boundaries from a PDB construct, which may carry tags or truncations.