Folklore

Folklore automates ACMG/AMP 2015 variant classification using a deterministic, rule-based Bayesian point-based framework with BayesDel ClinGen SVI-calibrated thresholds for clinical genetics laboratories.

Key Features:

• Bayesian Classification Framework: Implements Tavtigian et al. 2018 Bayesian point-based framework with Pejaver et al. 2022 BayesDel ClinGen SVI-calibrated thresholds for deterministic variant interpretation.
• Multi-Database Integration: Integrates 8 reference databases including gnomAD v4.1, ClinVar, dbNSFP 4.9c, SpliceAI, gnomAD Constraint, HPO, ClinGen, and Ensembl VEP.
• Comprehensive Variant Analysis: Analyzes nuclear and mtDNA variants, structural variants, and copy-number variants with trio/family and cohort analysis capabilities.
• Phenotype and Literature Integration: Supports HPO-based phenotype matching and biomedical literature mining across 2M+ PubMed publications.
• AI-Assisted Evidence Synthesis: AI assists in evidence synthesis but does not make classification decisions.
• Public MCP Adapter: Provides read-only Model Context Protocol adapter giving AI clients structured access to public germline variant evidence, automated ACMG/AMP decision support, provenance, and related literature without patient, phenotype, family, or case context.

Scientific Applications:

• Clinical Germline Variant Interpretation: Automates ACMG/AMP 2015 classification for clinical genetics laboratories processing germline variants.
• Structural and Copy-Number Variant Analysis: Supports interpretation of SV/CNV variants alongside single nucleotide variants.
• Family-Based Variant Analysis: Enables trio and family-based variant interpretation for inherited genetic conditions.
• Cohort-Level Variant Studies: Supports analysis across patient cohorts for population-scale variant interpretation.
• Mitochondrial DNA Variant Assessment: Analyzes mtDNA variants for mitochondrial disease diagnosis.
• Phenotype-Driven Variant Prioritization: Uses HPO terms to match patient phenotypes with variant evidence for diagnostic workflows.

Methodology:

Deterministic, rule-based interpretation using a Bayesian point-based framework with BayesDel ClinGen SVI-calibrated thresholds; integrates evidence from 8 reference databases including gnomAD v4.1, ClinVar, dbNSFP 4.9c, SpliceAI, gnomAD Constraint, HPO, ClinGen, and Ensembl VEP.