domains / science / clinical-variant-interpretation / clinical-interpretation

clinical-interpretation

Skill card

combined

The two skills cover the same task but with materially different and non-overlapping depth. The first copy (OpenClaw) is a practical, command-line workflow focused on downloading ClinVar, using bcftools for filtering, and running InterVar, but it relies on the older, flat 2015 ACMG criteria. The second copy (GPTomics) is a conceptual and methodological guide that explicitly warns against the 2015 defaults, mandates the use of 2018-2025 ClinGen refinements (e.g., graded PVS1, PM2 downgrade, PP5/BP6 retirement), and distinguishes between germline and somatic frameworks. Because the first copy's core classification method is explicitly identified as obsolete by the second, and the second lacks the specific shell commands for data retrieval and filtering found in the first, neither is a complete reproduction of the other. Combining them is necessary to provide a workflow that is both executable and methodologically current.

Repositories

Other copies

Combined skill

# Clinical Variant Interpretation

## 1. Framework Selection
Before interpreting, determine the variant origin:
- **Germline:** Use ACMG/AMP with ClinGen SVI refinements.
- **Somatic:** Use AMP/ASCO/CAP tiers or ClinGen/CGC/VICC oncogenicity. **Never** apply germline ACMG to somatic variants.

Check for a gene-specific ClinGen VCEP specification. If present, it overrides generic ACMG defaults.

## 2. Data Annotation
Use `bcftools` to annotate variants with ClinVar and gnomAD.

```bash
# Download ClinVar
wget https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/clinvar.vcf.gz
wget https://ftp.ncbi.nlm.nih.gov/pub/clinvar/vcf_GRCh38/clinvar.vcf.gz.tbi

# Annotate
bcftools annotate \
    -a clinvar.vcf.gz \
    -c INFO/CLNSIG,INFO/CLNDN,INFO/CLNREVSTAT \
    input.vcf.gz -Oz -o with_clinvar.vcf.gz
```

## 3. Filtering
Filter for rare variants and high-confidence ClinVar assertions.

```bash
# Rare variants (gnomAD AF < 0.01)
bcftools view -i 'INFO/gnomAD_AF<0.01 || INFO/gnomAD_AF="."' \
    with_clinvar.vcf.gz -Oz -o rare.vcf.gz

# High-confidence ClinVar (2+ stars)
bcftools view -i 'INFO/CLNREVSTAT~"multiple_submitters" || \
    INFO/CLNREVSTAT~"expert_panel" || \
    INFO/CLNREVSTAT~"practice_guideline"' \
    with_clinvar.vcf.gz -Oz -o high_confidence.vcf.gz
```

## 4. ACMG/AMP Classification (Germline)
Apply the **current** ClinGen SVI refinements, not the 2015 defaults:
- **PVS1:** Use the graded decision tree (Very Strong to Supporting) based on NMD and mechanism.
- **PM2:** Apply as **Supporting** (not Moderate).
- **PP5/BP6:** **Retired.** Do not use. Cite underlying data instead.
- **PP3/BP4:** Use one calibrated predictor at its calibrated strength.
- **Combining:** Use the Bayesian points system (Tavtigian 2020) for reproducible classification.

## 5. Automated Classification
Use InterVar for automated ACMG classification, ensuring it is configured to use the latest ClinGen specifications.

```bash
# Convert VCF to ANNOVAR format
convert2annovar.pl -format vcf4 input.vcf > input.avinput

# Run InterVar
python Intervar.py -i input.avinput -o intervar_results -b hg38
```

## 6. Interpretation
- **ClinVar is a lead, not evidence.** Re-derive from underlying data.
- **Context matters:** Pin the genome build, MANE Select transcript, and disease mechanism.
- **VUS Reanalysis:** Build a loop to re-evaluate VUS as new data becomes available.