tooluniverse-sequence-retrieval

This skill guides an agent through retrieving DNA, RNA, and protein sequences from the NCBI and ENA biological databases using the ToolUniverse toolkit. It solves a recurring problem in bioinformatics work: sequence identifiers are ambiguous, databases overlap, and naive lookups return low-quality or wrong-organism records. The skill encodes a disambiguation-first workflow and an explicit quality hierarchy (curated RefSeq NM_/NP_ accessions rank above predicted RefSeq XM_/XP_, which rank above raw GenBank submissions), so the agent fetches the best available sequence rather than the first one it finds.

The workflow runs through defined phases: optional clarification, gene and organism disambiguation via an accession-prefix decision tree, silent search-and-retrieve against NCBI Nucleotide (with ENA as a fallback for non-RefSeq accessions), and a structured Sequence Profile Report. It documents specific ToolUniverse tools (NCBI_search_nucleotide, NCBI_fetch_accessions, NCBI_get_sequence, ena_get_entry, ena_get_sequence_fasta), fallback chains, MANE Select transcript preference for human isoforms, curation-level tiers, error handling for common failures like ENA 404s on RefSeq accessions, and cross-database reconciliation (e.g. GenBank U00096 mapping to RefSeq NC_000913 for E. coli). Worked examples cover fetching the E. coli K-12 genome and the human BRCA1 gene.

It targets bioinformaticians, computational biologists, and researchers who need reproducible, correctly-disambiguated sequence retrieval for downstream BLAST, alignment, or annotation work. It is a read-only data-retrieval helper — it queries public reference databases and formats results; it does not modify data or run analyses itself.