---

name: score-rank description: Calculate final scores combining efficacy and safety

Score and Rank Candidates Skill

When to Use This Skill

Use after BLAST screening to compute final rankings combining design quality, gene essentiality, and safety.

Data Storage Structure

Reads from:

• output/{run}/candidates.json - From design-dsrna step
• output/{run}/blast_results.json - From blast-screen step
• output/{run}/essential_genes.json - From identify-genes step

Writes to:

• output/{run}/ranked_candidates.json - Final ranked list
• output/{run}/figures/ - Scoring visualization plots

Instructions

Step 1: Run Scoring Script

Copypython .deepagents/skills/score-rank/scripts/calculate_scores.py \
  --candidates output/{run}/candidates.json \
  --blast-results output/{run}/blast_results.json \
  --essential-genes output/{run}/essential_genes.json \
  --output output/{run}/ranked_candidates.json

Step 2: Verify Rankings

Copyjq '.[0:5] | .[] | {id, gene_name, combined_score, safety_status}' \
  output/{run}/ranked_candidates.json

Step 3: Generate Visualization

Create comprehensive scoring plots:

Copypython .deepagents/skills/score-rank/scripts/plot_rankings.py \
  --ranked output/{run}/ranked_candidates.json \
  --output-dir output/{run}/figures/

This creates:

• score_breakdown.png - Stacked bar chart showing efficacy/safety components
• efficacy_vs_safety_scatter.png - Scatter plot with candidates labeled
• top_candidates_radar.png - Radar chart comparing top 5 across all metrics

Step 4: Present Results

Output this summary to the user:

Copy## Score and Rank Complete

**Top 5 Candidates:**
| Rank | Candidate | Gene | Efficacy | Safety | Combined |
|------|-----------|------|----------|--------|----------|
| 1 | vATPase_1 | vATPase | 0.87 | 1.0 | 0.87 |
| 2 | ... | ... | ... | ... | ... |

**Top Recommendation:** {top_candidate} targeting {gene}
- Combined score: {score}
- Rationale: {why_this_candidate}

**Files Created:**
- `output/{run}/ranked_candidates.json`
- `output/{run}/figures/score_breakdown.png`

**Figures:** [Show efficacy vs safety scatter plot]

---
Proceed to generate-report? (yes/no)

Insights to include:

• Which genes appear multiple times in top 5
• Trade-offs between efficacy and safety
• Whether literature-supported genes rank highly

Scoring Formula

Efficacy Score (0-1)

Copyefficacy = 0.3×GC_score + 0.2×poly_n_score + 0.2×position_score + 0.3×gene_score

Safety Score (0-1)

Combined Score

Copycombined = efficacy × safety

Note: Rejected candidates (≥19bp match) get combined score of 0.

Output Format

output/{run}/ranked_candidates.json:

Copy[
  {
    "id": "vATPase_1",
    "gene_name": "vATPase",
    "gene_id": "lcl|NC_XXX",
    "sequence": "ATGCGT...",
    "start": 150,
    "end": 450,
    "length": 300,
    "gc_content": 0.423,
    "has_poly_n": false,
    "design_score": 5,
    "efficacy_score": 0.87,
    "safety_score": 1.0,
    "combined_score": 0.87,
    "human_max_match": 12,
    "honeybee_max_match": 14,
    "safety_status": "safe"
  }
]

Sorted by combined_score descending.

Expected Output

All outputs go in output/{run}/:

• ranked_candidates.json
• figures/score_breakdown.png
• figures/efficacy_vs_safety_scatter.png
• figures/top_candidates_radar.png

Available Tools

• shell - Run Python script and plotting
• read_file / write_file - Handle JSON

Interpretation Guide

High Combined Score (>0.7)

• Excellent candidate for synthesis
• Good GC content, no poly-N, essential gene target, safe

Medium Combined Score (0.4-0.7)

• Acceptable candidate
• May have suboptimal GC or caution-level off-targets

Low Combined Score (<0.4)

• Consider alternatives
• Either low efficacy or safety concerns