mtDNA variants

Each cell contains hundreds to thousands of mtDNA copies. A pathogenic variant must reach a sufficiently high fraction before energy production visibly suffers. This fraction, the heteroplasmy level, can vary strongly between tissues, so a blood finding does not represent all organs. Mutated and normal copies can also shift across generations and years of life, so the fraction and its effect are not static.

Genome calls mtDNA variants with bcftools, restricted to the mitochondrial contig (chrM, chrMT, or MT depending on the BAM). Concretely it runs bcftools mpileup on the region, then bcftools call with a haploid assumption (ploidy 1) and the multiallelic model, since mtDNA is circular and not diploid. The comparison uses the reference the BAM was aligned against; for hg38 the chrM contig is the rCRS. There is no dedicated heteroplasmy caller such as mutserve or Mutect2: Genome shows a heteroplasmy or depth value only when the VCF carries such a field, and otherwise marks it as not reported. The called variants are then placed against a curated mitochondrial disease context (Genome MitochondrialVariantDB, drawn from ClinVar, MITOMAP, and HmtVar).

The rCRS (NC_012920) is the standard reference against which mtDNA positions are named in m. notation; without this anchor, position calls are not comparable. A key source of error is NUMTs, nuclear-encoded mitochondrial pseudogenes whose reads can map wrongly to mtDNA and mimic apparent low heteroplasmy. The default haploid call works well for clear variants and for haplogroup preparation, but it is not designed to reliably separate genuine low heteroplasmy from NUMT or sequencing artifacts. Genome therefore presents the technical variant review, separates it from interpretation, and avoids diagnostic overclaims; robust statements about heteroplasmy require validated confirmation together with coverage and tissue context.