Chapter 10: ACMG Combining Rules, Bayesian Framework & Clinical Report Generation

Chapter 10: ACMG Combining Rules, Bayesian Framework & Clinical Report Generation

In this final capstone chapter, we bring every strand of the masterclass together. We master the formal ACMG criteria combination rules matrix, explore the mathematical Bayesian point system, solve the Variant of Uncertain Significance (VUS) dilemma, and engineer an automated clinical diagnostic report pipeline.

1. ACMG Combining Rules for Final Verdict Matrix

2. The Bayesian Point Framework (Tavtigian / ClinGen)

To remove arbitrary threshold boundaries, modern computational genetics formulates ACMG rules as an exponential Bayesian point scale:

• PVS1 = +8 Points (Odds of Pathogenicity $O_P = 350:1$)
• PS1–PS4 = +4 Points each ($O_P = 18.7:1$)
• PM1–PM6 = +2 Points each ($O_P = 4.3:1$)
• PP1–PP5 = +1 Point each ($O_P = 2.08:1$)
• BP1–BP7 = -1 Point each
• BS1–BS4 = -4 Points each
• BA1 = Stand-Alone Benign ($-\infty$)

3. Automated Clinical Genetics Report Generator in Python

Below is a production-grade Python implementation of the ACMG combining rules engine that formats a final clinical diagnostic report:

class ACMGClassifier:
    def __init__(self, gene: str, hgvs_c: str, hgvs_p: str, criteria: list):
        self.gene = gene
        self.hgvs_c = hgvs_c
        self.hgvs_p = hgvs_p
        self.criteria = set([c.upper().strip() for c in criteria])
        
    def classify(self) -> str:
        c = self.criteria
        
        # 1. Stand-Alone Benign Check
        if 'BA1' in c or sum(1 for x in c if x.startswith('BS')) >= 2:
            return "BENIGN"
            
        # Count pathogenic evidence weights
        pvs = 1 if 'PVS1' in c else 0
        ps  = sum(1 for x in c if x.startswith('PS'))
        pm  = sum(1 for x in c if x.startswith('PM'))
        pp  = sum(1 for x in c if x.startswith('PP'))
        
        # Count benign evidence weights
        bs  = sum(1 for x in c if x.startswith('BS'))
        bp  = sum(1 for x in c if x.startswith('BP'))
        
        # Check for Pathogenic
        if (pvs == 1 and ps >= 1) or            (pvs == 1 and pm >= 2) or            (pvs == 1 and pm == 1 and pp >= 1) or            (pvs == 1 and pp >= 2) or            (ps >= 2) or            (ps == 1 and pm >= 3) or            (ps == 1 and pm == 2 and pp >= 2) or            (ps == 1 and pm == 1 and pp >= 4):
            return "PATHOGENIC"
            
        # Check for Likely Pathogenic
        if (pvs == 1 and pm == 1) or            (ps == 1 and 1 <= pm <= 2) or            (ps == 1 and pp >= 2) or            (pm >= 3) or            (pm == 2 and pp >= 2) or            (pm == 1 and pp >= 4):
            return "LIKELY PATHOGENIC"
            
        # Check for Likely Benign
        if (bs >= 1 and bp >= 1) or (bp >= 2):
            return "LIKELY BENIGN"
            
        return "VARIANT OF UNCERTAIN SIGNIFICANCE (VUS)"

    def generate_report(self, patient_id: str) -> str:
        verdict = self.classify()
        return f"""========================================================================================
                      CLINICAL MOLECULAR GENETICS LABORATORY REPORT
========================================================================================
PATIENT ID:       {patient_id}
PRIMARY FINDING:  {self.gene} ({self.hgvs_c}, {self.hgvs_p})
ACMG CRITERIA:    {', '.join(sorted(self.criteria))}
FINAL DIAGNOSIS:  {verdict}
----------------------------------------------------------------------------------------
CLINICAL EVIDENCE RATIONALE:
- The variant was analyzed following ACMG/AMP 2015 and ClinGen recommendations.
- Functional and population database evidence triggered criteria: {', '.join(sorted(self.criteria))}.
- Final diagnostic verdict calculated as {verdict}.
========================================================================================"""

# Example Execution: BRCA1 Frameshift Variant
classifier = ACMGClassifier("BRCA1", "c.5266dupC", "p.Gln1756Profs*74", ["PVS1", "PM2", "PS4"])
print(classifier.generate_report("PATIENT_OMICS_8921"))