Mitochondrial Variant Analysis¶

Classify mitochondrial DNA (chrM/MT) variants using mtDNA-specific criteria distinct from the nuclear ACMG/AMP 2015 framework. Automatic detection, no special flags required.

Quick start¶

# mtDNA variants are automatically detected from chrM/MT in the VCF
vartriage --vcf wgs.vcf.gz --output results.json

# Override the minimum heteroplasmy threshold (default: 1%)
vartriage --vcf wgs.vcf.gz --output results.json --mt-min-heteroplasmy 5.0

# Skip mitochondrial analysis (targeted panels without mtDNA capture)
vartriage --vcf panel.vcf.gz --output results.json --skip-mito

How it works¶

When the pipeline encounters chrM/MT variants in the input VCF, it routes them to a separate mitochondrial sub-pipeline while nuclear variants proceed through the standard ACMG flow. Results from both pipelines merge at the report stage.

VCFParser
  ├── chrM/MT variants → MitochondrialPipeline
  │     HeteroplasmyExtractor → GeneMapAnnotator → MITOMAP/HelixMTdb → Classifier
  │
  └── Nuclear variants → Standard ACMG Pipeline
        QualityFilter → AnnotationEngine → Prioritization → ACMGClassifier

Both streams → ReportGenerator (merged output)

Mitochondrial genetic code¶

The vertebrate mitochondrial genome uses a different translation table at 4 codons:

Codon	Standard Code	Mitochondrial Code
TGA	Stop (*)	Trp (W)
ATA	Ile (I)	Met (M)
AGA	Arg (R)	Stop (*)
AGG	Arg (R)	Stop (*)

VarTriage automatically selects the mitochondrial translation table when resolving amino acid changes for chrM/MT variants. This prevents false stop-gain calls (e.g., a TGA codon in MT-ATP6 is Trp, not a premature stop).

Heteroplasmy¶

Mitochondrial DNA exists in hundreds to thousands of copies per cell. A variant can be present in a fraction of copies rather than the binary het/hom states of nuclear diploid variants.

Extraction¶

Heteroplasmy is extracted from VCF FORMAT fields:

AD field (primary): ALT_depth / (REF_depth + ALT_depth)
AF field (fallback): Used directly when AD is unavailable (e.g., Mutect2 mitochondrial mode)

Level classification¶

Category	Range	Clinical relevance
Homoplasmic	>= 95%	Functionally equivalent to fixed mutation
High	60-95%	Strong phenotypic effect expected
Moderate	20-60%	Variable expressivity, tissue-dependent
Low	1-20%	Usually subclinical, monitor over time
Sub-threshold	< 1%	Likely sequencing noise, filtered by default

Filtering¶

Variants below the --mt-min-heteroplasmy threshold (default 1.0%) are excluded from output. This filters sequencing artifacts and ultra-low-level noise.

Classification criteria¶

mtDNA variants are classified using a rule-based system independent of ACMG/AMP 2015:

Pathogenic¶

All three conditions must be met:

Confirmed pathogenic in MITOMAP (status = "Cfrm")
High or homoplasmic heteroplasmy (>= 60%)
Rare in HelixMTdb (AF < 0.01%)

Likely Pathogenic¶

Reported in MITOMAP (any status) AND
Located in a protein-coding gene or tRNA AND
Moderate or higher heteroplasmy (>= 20%)

Also assigned when a variant is confirmed in MITOMAP and rare, but heteroplasmy data is unavailable.

Benign¶

Common haplogroup-defining polymorphism: AF > 5% in HelixMTdb

Likely Benign¶

Moderate population frequency: AF > 0.1% in HelixMTdb

VUS (Variant of Uncertain Significance)¶

Default for variants without strong evidence in either direction
Includes novel variants absent from both MITOMAP and HelixMTdb

Data sources¶

MITOMAP¶

MITOMAP is the primary mtDNA disease database. The bundled dataset includes ~60 confirmed and reported pathogenic mutations covering:

MELAS (m.3243A>G, m.3271T>C)
MERRF (m.8344A>G, m.8356T>C)
LHON (m.3460G>A, m.11778G>A, m.14484T>C)
NARP/Leigh syndrome (m.8993T>G/C)
Aminoglycoside-induced deafness (m.1555A>G)

Update the bundled data:

python scripts/download_mitomap.py --output vartriage/data/mito/mitomap_pathogenic.tsv

HelixMTdb¶

Population allele frequencies from HelixMTdb, a large-scale mitochondrial database (~200k samples). Used to distinguish common haplogroup markers from rare potentially pathogenic variants.

Update the bundled data:

python scripts/download_helixmtdb.py --output vartriage/data/mito/helixmtdb_frequency.tsv

MT gene map¶

The bundled gene map covers all 37 mitochondrial genes:

13 protein-coding genes (MT-ND1 through MT-ND6, MT-ND4L, MT-CO1-3, MT-ATP6, MT-ATP8, MT-CYB)
22 tRNA genes (MT-TF, MT-TL1, MT-TL2, MT-TK, etc.)
2 rRNA genes (MT-RNR1, MT-RNR2)
D-loop/control region (positions 1-576 and 16024-16569)

Coordinates follow the revised Cambridge Reference Sequence (rCRS, NC_012920).

Maternal inheritance check¶

When trio data is available (proband + mother + father), the pipeline verifies maternal inheritance:

Maternal: Present in mother, absent in father (expected for mtDNA)
De novo: Absent in both parents (rare but clinically significant)
Paternal unexpected: Detected in father (indicates NUMTs, contamination, or data error)

This check is automatic when inheritance analysis is configured with --proband, --mother, and --father flags.

Output format¶

JSON output¶

When mitochondrial variants are present, JSON output includes a mitochondrial_findings array:

{
  "variants": [...],
  "mitochondrial_findings": [
    {
      "chromosome": "chrM",
      "position": 3243,
      "ref_allele": "A",
      "alt_allele": "G",
      "mt_classification": "Likely_Pathogenic",
      "heteroplasmy_level": 85.2,
      "heteroplasmy_category": "high",
      "heteroplasmy_depth": 1200,
      "mitomap_disease": "MELAS / Leigh syndrome / Diabetes-Deafness",
      "mitomap_status": "Cfrm",
      "mt_gene": "MT-TL1",
      "mt_gene_type": "tRNA",
      "helix_af": 0.0002,
      "classification_reason": "Reported in MITOMAP ..."
    }
  ],
  "metadata": {
    "mitochondrial_note": "Mitochondrial variants classified using mtDNA-specific criteria"
  }
}

CSV output¶

CSV output appends a mitochondrial findings section after the nuclear variants with dedicated columns:

mt_classification, heteroplasmy_level, heteroplasmy_category
mitomap_disease, mitomap_status
mt_gene, mt_gene_type, helix_af

Clinical reports¶

Clinical reports (HTML/PDF/DOCX) include a "Mitochondrial Findings" section with a table showing position, nucleotide change, gene, classification, heteroplasmy level, and MITOMAP disease association.

CLI reference¶

Flag	Default	Description
`--skip-mito`	disabled	Skip mitochondrial analysis entirely
`--mt-min-heteroplasmy`	1.0	Minimum heteroplasmy % for reporting (0.0-100.0)

Module structure¶

vartriage/mito/
├── __init__.py          # Package marker
├── genetic_code.py      # MT codon table + is_mitochondrial() helper
├── gene_map.py          # 37-gene interval lookup (binary search)
├── heteroplasmy.py      # AD/AF extraction + level classification
├── mitomap.py           # MITOMAP disease database (O(1) lookup)
├── frequency.py         # HelixMTdb population frequency (O(1) lookup)
├── classifier.py        # 5-rule classification decision tree
├── maternal.py          # Trio maternal inheritance verification
├── pipeline.py          # Sub-pipeline orchestrator
├── config.py            # MitoConfig dataclass
└── report.py            # Output serialization for JSON/CSV

vartriage/data/mito/
├── mt_gene_map.tsv          # Gene intervals (rCRS coordinates)
├── mitomap_pathogenic.tsv   # Curated pathogenic mutations
└── helixmtdb_frequency.tsv  # Population allele frequencies