Clinical InterpretationSAFE

The largest open-source medical AI skills library for OpenClaw🦞.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

<!--
# COPYRIGHT NOTICE
# This file is part of the "Universal Biomedical Skills" project.
# Copyright (c) 2026 MD BABU MIA, PhD <[email protected]>
# All Rights Reserved.
#
# This code is proprietary and confidential.
# Unauthorized copying of this file, via any medium is strictly prohibited.
#
# Provenance: Authenticated by MD BABU MIA

-->

---
name: bio-variant-calling-clinical-interpretation
description: Clinical variant interpretation using ClinVar, ACMG guidelines, and pathogenicity predictors. Prioritize variants for diagnostic and research applications. Use when interpreting clinical significance of variants.
tool_type: mixed
primary_tool: InterVar
measurable_outcome: Execute skill workflow successfully with valid output within 15 minutes.
allowed-tools:
  - read_file
  - run_shell_command
---

# Clinical Variant Interpretation

Prioritize and interpret variants for clinical significance using databases and ACMG/AMP guidelines.

## Interpretation Framework

```
Annotated VCF
    │
    ├── Database Lookup
    │   ├── ClinVar (clinical assertions)
    │   ├── OMIM (disease associations)
    │   └── gnomAD (population frequency)
    │
    ├── Computational Predictions
    │   ├── SIFT, PolyPhen-2
    │   ├── CADD, REVEL
    │   └── SpliceAI
    │
    ├── ACMG Classification
    │   └── Pathogenic → Likely Pathogenic → VUS → Likely Benign → Benign
    │
    └── Prioritized Variant List
```

## ClinVar Annotation

### Download ClinVar

```bash
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 with bcftools

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

### Filter Pathogenic Variants

```bash
# Pathogenic or Likely pathogenic
bcftools view -i 'INFO/CLNSIG~"Pathogenic" || INFO/CLNSIG~"Likely_pathogenic"' \
    with_clinvar.vcf.gz -Oz -o pathogenic.vcf.gz

# Exclude benign
bcftools view -e 'INFO/CLNSIG~"Benign" || INFO/CLNSIG~"Likely_benign"' \
    with_clinvar.vcf.gz -Oz -o not_benign.vcf.gz
```

## ClinVar Significance Levels

| CLNSIG | Meaning | Action |
|--------|---------|--------|
| Pathogenic | Disease-causing | Report |
| Likely_pathogenic | Probably disease-causing | Report with caveat |
| Uncertain_significance | VUS | May report, needs follow-up |
| Likely_benign | Probably not disease-causing | Usually exclude |
| Benign | Not disease-causing | Exclude |
| Conflicting | Multiple interpretations | Manual review |

## ClinVar Review Status

| CLNREVSTAT | Stars | Meaning |
|------------|-------|---------|
| practice_guideline | 4 | Expert panel reviewed |
| reviewed_by_expert_panel | 3 | ClinGen expert reviewed |
| criteria_provided,_multiple_submitters | 2 | Consistent assertions |
| criteria_provided,_single_submitter | 1 | One submitter with criteria |
| no_assertion_criteria | 0 | No criteria provided |

```bash
# Filter for high-confidence assertions (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
```

## InterVar (ACMG Classification)

Automated ACMG/AMP variant classification.

### Installation

```bash
git clone https://github.com/WGLab/InterVar.git
cd InterVar
# Download databases per documentation
```

### Run InterVar

```bash
python Intervar.py \
    -i input.avinput \
    -o output \
    -b hg38 \
    -d humandb/ \
    --input_type=AVinput
```

### From VCF

```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
```

## ACMG/AMP Criteria

### Pathogenic Criteria

| Code | Type | Description |
|------|------|-------------|
| PVS1 | Very Strong | Null variant in gene where LOF is disease mechanism |
| PS1-4 | Strong | Same AA change, functional studies, etc. |
| PM1-6 | Moderate | Hot spot, absent from controls, etc. |
| PP1-5 | Supporting | Co-segregation, computational evidence |

### Benign Criteria

| Code | Type | Description |
|------|------|-------------|
| BA1 | Stand-alone | AF >5% in gnomAD |
| BS1-4 | Strong | AF greater than expected, functional studies |
| BP1-7 | Supporting | Missense in gene with truncating mechanism |

## Population Frequency Filtering

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

# Ultra-rare for dominant diseases (AF < 0.0001)
bcftools view -i 'INFO/gnomAD_AF<0.0001 || INFO/gnomAD_AF="."' \
    input.vcf.gz -Oz -o ultrarare.vcf.gz
```

## Pathogenicity Score Filtering

### CADD Scores

```bash
# CADD > 20 (top 1% deleterious)
bcftools view -i 'INFO/CADD_PHRED>20' input.vcf.gz -Oz -o cadd_filtered.vcf.gz

# CADD > 30 (top 0.1%)
bcftools view -i 'INFO/CADD_PHRED>30' input.vcf.gz -Oz -o highly_deleterious.vcf.gz
```

### REVEL Scores

```bash
# REVEL > 0.5 (likely pathogenic)
bcftools view -i 'INFO/REVEL>0.5' input.vcf.gz -Oz -o revel_filtered.vcf.gz
```

### Combined Filtering

```bash
bcftools view -i '(INFO/CADD_PHRED>20 || INFO/REVEL>0.5) && \
    (INFO/CLNSIG~"Pathogenic" || INFO/CLNSIG~"Likely" || INFO/CLNSIG=".")' \
    input.vcf.gz -Oz -o prioritized.vcf.gz
```

## Python: Clinical Prioritization

```python
from cyvcf2 import VCF, Writer

def classify_variant(variant):
    clnsig = variant.INFO.get('CLNSIG', '')
    af = variant.INFO.get('gnomAD_AF', 0) or 0
    cadd = variant.INFO.get('CADD_PHRED', 0) or 0
    revel = variant.INFO.get('REVEL', 0) or 0

    # Known pathogenic
    if 'Pathogenic' in str(clnsig):
        return 'PATHOGENIC'
    if 'Likely_pathogenic' in str(clnsig):
        return 'LIKELY_PATHOGENIC'

    # Known benign
    if 'Benign' in str(clnsig) or af > 0.05:
        return 'BENIGN'

    # Computational predi

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (1)

Gates applied: no_behavioural_pass.