---

name: fetch-genome description: Download CDS sequences from NCBI RefSeq for a target species

Fetch Genome Skill

When to Use This Skill

Use at the start of dsRNA design workflow when you need coding sequences (CDS) for a pest species from NCBI.

Data Storage Structure

Input data (cached, reusable) goes in data/{assembly}/:

• genome.fasta - Downloaded CDS sequences
• genome_metadata.json - Assembly info, download date

Analysis outputs go in output/{run}/ where {run} is created by the full-workflow skill as YYYYMMDD-HHMMSS-{species_slug}.

IMPORTANT: Check Cache First

Before downloading, always check if the genome data already exists:

Copyls data/*/genome_metadata.json 2>/dev/null | head -5

If you find a matching assembly for your species, skip the download and use the existing data. Read the metadata to confirm it's the right species:

Copycat data/{assembly}/genome_metadata.json

Instructions

Step 1: Resolve Species to TaxID

Use fetch_url to query NCBI Taxonomy:

Copyhttps://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=taxonomy&term={species}&retmode=json

Parse JSON response: esearchresult.idlist[0] is the TaxID

Step 2: Find RefSeq Assembly

Use fetch_url:

Copyhttps://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=assembly&term=txid{TAXID}[Organism:exp]+AND+latest_refseq[filter]&retmode=json

Get assembly ID from esearchresult.idlist[0]

Step 3: Get Assembly Accession and FTP Path

Use fetch_url:

Copyhttps://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi?db=assembly&id={ASSEMBLY_ID}&retmode=json

Extract from response:

• Assembly accession: result[assembly_id].assemblyaccession (e.g., "GCF_000001215.4")
• FTP path: result[assembly_id].ftppath_refseq

Step 4: Check if Already Downloaded

CRITICAL: Check if this assembly is already cached:

Copyif [ -f "data/{ASSEMBLY_ACCESSION}/genome.fasta" ]; then
    echo "CACHED: Genome already exists"
else
    echo "NOT CACHED: Need to download"
fi

If cached, skip to Step 8 (literature search).

Step 5: Create Data Directory

Copymkdir -p data/{ASSEMBLY_ACCESSION}
mkdir -p data/{ASSEMBLY_ACCESSION}/figures

Use the assembly accession (e.g., "GCF_000001215.4") as the folder name.

Step 6: Download CDS FASTA

Use execute to download:

CopyFTP_PATH="https://ftp.ncbi.nlm.nih.gov/..."  # from step 3
ASSEMBLY_NAME=$(basename $FTP_PATH)
curl -L -o data/{ASSEMBLY_ACCESSION}/genome.fasta.gz \
  "${FTP_PATH}/${ASSEMBLY_NAME}_cds_from_genomic.fna.gz"
gunzip data/{ASSEMBLY_ACCESSION}/genome.fasta.gz

If _cds_from_genomic.fna.gz returns 404, try _rna.fna.gz instead.

Step 7: Verify Download

Use execute:

Copygrep -c "^>" data/{ASSEMBLY_ACCESSION}/genome.fasta

Should show thousands of sequences (typically 10,000-20,000 for insects)

Step 8: Save Metadata

Use write_file to save data/{ASSEMBLY_ACCESSION}/genome_metadata.json:

Copy{
  "species": "{species_name}",
  "taxid": "{TAXID}",
  "assembly_id": "{ASSEMBLY_ID}",
  "assembly_accession": "{ASSEMBLY_ACCESSION}",
  "ftp_path": "{FTP_PATH}",
  "download_date": "{ISO_DATE}",
  "sequence_count": {COUNT},
  "total_length": {LENGTH}
}

Step 9: Search Literature (Automatic)

NOTE: Literature search results are analysis outputs, NOT cached input data. Save them to output/{run}/literature_search.json, not in data/.

Automatically search PubMed for RNAi studies on this species:

Copypubmed_search_articles
query: "{species}" AND (RNAi OR dsRNA OR "RNA interference" OR "gene silencing")
max_results: 50

IMPORTANT: When saving results, you MUST extract gene names from each paper's title and abstract. See dsrna_agent/skills/literature-search/SKILL.md for the list of gene patterns to look for (vATPase, chitin synthase, acetylcholinesterase, etc.)

Save to output/{run}/literature_search.json in this format:

Copy[
  {
    "pmid": "12345678",
    "title": "...",
    "gene_names": ["vATPase", "chitin synthase"],
    ...
  }
]

The gene_names field is REQUIRED for downstream scripts to give literature support scores to candidate genes.

This step does NOT require user confirmation - literature search is automatic.

Step 10: Present Results

Output this summary to the user:

Copy## Fetch Genome Complete

**Species:** {species_name} (TaxID: {taxid})
**Assembly:** {assembly_accession}
**Status:** {DOWNLOADED or CACHED}

**Summary:**
- {sequence_count} CDS sequences
- Total length: {total_length} bp

**Literature:** Found {paper_count} RNAi papers, top genes: {gene_list}

**Cached Input Data:**
- `data/{assembly_accession}/genome.fasta`
- `data/{assembly_accession}/genome_metadata.json`

**Analysis Output:**
- `output/{run}/literature_search.json`

---
Proceed to identify-genes? (yes/no)

Expected Output

Cached input data (in data/{assembly}/):

• genome.fasta - Downloaded CDS sequences
• genome_metadata.json - TaxID, assembly ID, stats

Analysis outputs (in output/{run}/):

• literature_search.json - PubMed search results with extracted gene names

Available Tools

• fetch_url - Query NCBI APIs
• execute - Run curl/gunzip commands
• write_file - Save metadata
• pubmed_search_articles - Search literature