sequence-basic-analysis

name	sequence_basic_analysis
description	Analyze DNA/RNA/protein sequences and optionally probe UniProt for homology. Use when user provides FASTA format or raw sequence.
category	sequence
safety_level	safe
tools_used	["sequence.analyze","uniprot.search"]

Sequence Basic Analysis

Analyze biological sequence: $ARGUMENTS

Input

• sequence (required): FASTA format or raw DNA/RNA/protein sequence

Output

Returns structured analysis:

• annotation: Sequence metadata (kind, type, length)
• confidence: 0.7-0.85 depending on type
• sequence_stats: Detailed statistics per record
• uniprot_hint: Optional homology probe results (protein ≥25aa)

Workflow

Step 1: Parse and Analyze Sequence

Call sequence.analyze tool (local BioPython):

from open_rosalind.skills_v2.sequence import tools

stats = tools.analyze(sequence)
# Returns: {records: [{type, length, gc_content, translation, ...}], n_records}

For each record:

• Detect type (DNA/RNA/protein)
• Compute length, composition
• DNA: GC%, translation, reverse complement
• Protein: molecular weight

Step 2: Optional UniProt Probe (Protein ≥25aa)

If primary sequence is protein and length ≥25aa:

1. Extract probe fragment (30aa or full sequence if shorter)
2. Call uniprot.search with probe
3. Aggregate top hits into uniprot_hint

probe = primary_sequence[:30]
result = uniprot_search(query=probe)
uniprot_hint = {
    "hits": len(result["hits"]),
    "top_match": result["hits"][0]["accession"],
    "probe_length": 30
}

Step 3: Return Evidence-Grounded Result

Construct response following Open-Rosalind principles:

• annotation: Structured metadata (no free-form text)
• confidence: Rule-based score (0.85 for DNA/RNA, 0.7 for protein)
• notes: Fallback messages (e.g., "UniProt probe: 3 hits")
• sequence_stats: Full tool output (evidence)
• uniprot_hint: Probe results (evidence)

Examples

Example 1: DNA Sequence

Input:

{"sequence": "ATGGCCAAATTAA"}

Output:

{
  "annotation": {"kind": "sequence", "primary_type": "dna", "length": 13},
  "confidence": 0.85,
  "sequence_stats": {
    "records": [{
      "type": "dna",
      "length": 13,
      "gc_content": 30.77,
      "translation": "MAKL*",
      "reverse_complement": "TTAATTTGGCCAT"
    }]
  }
}

Example 2: Protein with UniProt Probe

Input:

{"sequence": "MVKVGVNGFGRIGRLVTRA"}

Output:

{
  "annotation": {"kind": "sequence", "primary_type": "protein", "length": 19},
  "confidence": 0.7,
  "notes": ["UniProt probe: 3 hits"],
  "sequence_stats": {
    "records": [{
      "type": "protein",
      "length": 19,
      "molecular_weight": 1.95
    }]
  },
  "uniprot_hint": {
    "hits": 3,
    "top_match": "Q9H3P7",
    "probe_length": 19
  }
}

Design Principles

Following Open-Rosalind core principles:

1. Tool-first: All facts from sequence.analyze and uniprot.search tools
2. Evidence-grounded: sequence_stats and uniprot_hint are raw tool outputs
3. Traceable: Every tool call logged to trace
4. Fail-safe: UniProt probe failure doesn't crash skill (logged to notes)

Notes

• Multi-FASTA supported (splits on > headers)
• UniProt probe only for protein ≥25aa (avoids false positives)
• Probe uses 30aa fragment (balance between specificity and coverage)
• Local computation (BioPython) is fast and deterministic