classify

name: classify description: Full ACMG/AMP variant classification workflow. Validates input, gathers evidence, applies rules, and generates classification. Invoked via /classify .

Variant Classification Workflow

Perform a complete ACMG/AMP variant classification using the MCP server tools.

Usage

/classify NM_000492.3:c.1521_1523delCTT
/classify BRCA1:c.5266dupC
/classify TP53 p.R273H
/classify chr17:g.43094692G>A --report

Supported Input Formats

Workflow Steps

Step 1: Input Validation

Use the validate_hgvs MCP tool to validate and enrich the input:

Tool: validate_hgvs
Parameters:
  - hgvs_notation: <user input>
  - strict_mode: false

Enhanced output includes:

• is_valid: Validation result
• normalized_hgvs: Standardized notation
• gene_info: Gene symbol, name, and HGNC ID
• transcript_info: RefSeq, Ensembl IDs, canonical status
• suggestions: Fix suggestions for invalid input

If validation fails:

1. Report the specific validation error
2. Present suggestions from the tool response
3. Reference .claude/skills/_shared/error-handling.md for common issues

If validation succeeds:

• Use normalized_hgvs for subsequent steps
• Display gene_info and transcript_info to user
• Proceed to evidence gathering

Step 2: Evidence Gathering

Use the query_evidence MCP tool to gather comprehensive evidence:

Tool: query_evidence
Parameters:
  - hgvs_notation: <normalized HGVS>
  - databases: ["clinvar", "gnomad", "cosmic", "pubmed"]

Enhanced output includes:

• database_results: Raw data from each source
• source_quality: Per-source quality assessment (high/medium/low)
• quality_scores: Overall evidence quality with completeness score
• acmg_criteria_hints: Pre-mapped ACMG criteria suggestions
• synthesis: Human-readable evidence summary

Present evidence summary:

• Use the synthesis field for a quick overview
• Highlight acmg_criteria_hints that apply
• Note data quality from source_quality

Step 3: Variant Classification

Use the classify_variant MCP tool for full classification:

Tool: classify_variant
Parameters:
  - hgvs_notation: <normalized HGVS> (or)
  - gene_symbol_notation: <gene:variant>
  - clinical_context: <if provided by user>

Classification analysis:

1. Review each ACMG/AMP rule applied
2. Cross-reference with acmg_criteria_hints from evidence
3. Explain evidence combination logic
4. Report final classification with confidence

Step 4: Results Summary

Present results in this format:

## Classification Result

**Variant:** [Normalized HGVS notation]
**Gene:** [Gene symbol] ([Transcript]) - from validate_hgvs gene_info
**Classification:** [Pathogenic/Likely Pathogenic/VUS/Likely Benign/Benign]
**Confidence:** [High/Medium/Low]

### Evidence Summary

[Include synthesis text from query_evidence]

| Criterion | Met | Strength | Evidence |
|-----------|-----|----------|----------|
| PVS1 | Yes/No | Very Strong | [Brief explanation] |
| PS1 | Yes/No | Strong | [Brief explanation] |
| ... | ... | ... | ... |

### Key Findings

- **Population Frequency:** [from query_evidence]
- **Clinical Assertions:** [from query_evidence]
- **Functional Impact:** [Computational predictions]

### Classification Rationale

[Explanation of how evidence was combined to reach classification]

Step 5 (Optional): Report Generation

If --report flag is provided, generate a clinical report:

Tool: generate_report
Parameters:
  - variant: <normalized HGVS>
  - template: "clinical"
  - include_evidence: true

Options

Error Handling

Refer to .claude/skills/_shared/error-handling.md for:

• Input validation errors
• Database connection issues
• Insufficient evidence scenarios

Clinical Guidelines

Refer to .claude/skills/_shared/acmg-guidelines.md for:

• ACMG/AMP rule definitions
• Evidence combination tables
• Classification thresholds

Examples

Example 1: CFTR Variant

User: /classify NM_000492.3:c.1521_1523delCTT

Claude:
## Classification Result

**Variant:** NM_000492.3:c.1521_1523delCTT (p.Phe508del)
**Gene:** CFTR (CF transmembrane conductance regulator)
**Transcript:** NM_000492.3 (canonical)
**Classification:** Pathogenic
**Confidence:** High

### Evidence Summary

Population frequency: absent/very rare - supports pathogenicity.
Clinical significance: Pathogenic with expert panel review.
Evidence quality: Excellent (95% complete).

| Criterion | Met | Strength | Evidence |
|-----------|-----|----------|----------|
| PS1 | Yes | Strong | Well-established pathogenic variant |
| PS4 | Yes | Strong | Significantly increased in CF patients |
| PP5 | Yes | Supporting | Multiple ClinVar pathogenic assertions |

### Classification Rationale

This variant meets criteria for Pathogenic classification based on:
2 Strong criteria (PS1, PS4) + 1 Supporting (PP5)

Example 2: Novel VUS

User: /classify BRCA1:c.1234A>G

Claude:
## Classification Result

**Variant:** NM_007294.4:c.1234A>G (p.Asn412Asp)
**Gene:** BRCA1 (BRCA1 DNA repair associated)
**Classification:** Uncertain Significance (VUS)
**Confidence:** Medium

### Evidence Summary

Population frequency: absent from gnomAD.
Computational prediction: damaging (score: 0.72).
Evidence quality: Moderate (65% complete).

| Criterion | Met | Strength | Evidence |
|-----------|-----|----------|----------|
| PM2 | Yes | Moderate | Absent from gnomAD |
| PP3 | Yes | Supporting | REVEL score 0.72 (damaging) |

### Classification Rationale

This variant is classified as VUS due to:
- Only 1 Moderate (PM2) + 1 Supporting (PP3) criteria met
- Insufficient evidence to reach Likely Pathogenic threshold

MCP Tools Used

References

• .claude/skills/_shared/acmg-guidelines.md - ACMG/AMP criteria reference
• .claude/skills/_shared/error-handling.md - Error handling guide
• .claude/skills/_shared/clinical-context.md - Clinical interpretation guidance