This study evaluated whether human-derived reads that are typically discarded in shotgun metagenomic pipelines can be repurposed to recover complete mitochondrial genomes and to detect mitochondrial heteroplasmy. The analysis used shotgun metagenomic data from 47 upper respiratory tract swabs collected from healthy adults in Nagpur, India. The authors report complete mitochondrial genome recovery in all samples, with mean coverage of 433X and a range of 9–3,029X, demonstrating that host reads in low-biomass clinical samples can yield substantive mtDNA sequence information.
Human reads were first recruited against the revised Cambridge Reference Sequence (rCRS) for the mitochondrial genome. To reduce false positive mitochondrial calls arising from nuclear mitochondrial segments, the recruited reads were realigned competitively to the GRCh38 nuclear reference; this step relocated reads that better matched nuclear loci (NUMTs). Kraken2 was used as a microbial control to identify and remove reads that could cross-map from microbial sequences.
Heteroplasmy calling used a pooled, substitution-specific beta-binomial error model combined with a leave-one-out panel of normals to model background noise. The authors evaluated the detection limit by comparing the observed substitution spectrum against gnomAD v3.1.
Across the 47 samples the mean mitochondrial sequencing depth was 433X (range 9–3,029X). Competitive realignment to GRCh38 reassigned 3.22% of reads originally recruited to the mitochondrial reference back to nuclear loci, indicating the presence of NUMT-related signal. Microbial cross-mapping was minimal, removing 0.045% of recruited reads. These statistics quantify the extent of non-mitochondrial contamination and illustrate the impact of competitive alignment and microbial filtering on the host-read dataset.
A total of 134 heteroplasmic variant calls were reported. The study found pronounced variation in error rates across substitution classes — a 12.8-fold difference — with T>A substitutions having the highest apparent error rate. The 134 calls had a transition/transversion (Ti/Tv) ratio of 0.43, markedly lower than the Ti/Tv of 3.51 observed in gnomAD v3.1. Notably, T>A substitutions accounted for 42.5% of calls and were concentrated at low allele fractions.
T>A allele fractions in the dataset never exceeded 4.72%, whereas transition-type variants reached allele fractions up to 90.81%. The authors interpret this as evidence of a substitution-specific background process that remains after NUMT removal and is not fully captured by the beta-binomial background model.
To explore detection limits, the authors incrementally raised the minimum allele-fraction threshold for heteroplasmy calls. At a 5% allele-fraction threshold, T>A calls disappeared from the call set and the Ti/Tv ratio increased to 2.20 before plateauing at higher thresholds. Sixteen heteroplasmic calls exceeded the 5% threshold. From these observations the authors infer that reliable heteroplasmy detection from metagenomic host reads is achievable for allele fractions greater than 5%, while calls below this threshold are dominated by a residual T>A error process not eliminated by NUMT realignment or the background model.
Macrohaplogroup assignment was possible for 46 of 47 samples. The distribution of macrohaplogroups in the study cohort was M (22 samples), R (23 samples), and N (1 sample). These assignments indicate that the recovered mitochondrial genomes were of sufficient quality for population-level haplogroup typing in nearly all samples.
The authors conclude that repurposing host-derived reads from shotgun metagenomic datasets can reliably recover complete mitochondrial genomes and support haplogroup typing. For heteroplasmy detection, metagenomic host reads are reliable for allele fractions above 5%. At allele fractions below 5%, a T>A-dominated error process persists despite NUMT removal and background fitting, limiting confident detection of low-level heteroplasmy in this data type. These findings suggest practical allele-fraction thresholds and highlight substitution-specific error behavior that must be considered when using host reads from metagenomic studies for mitochondrial variant analysis.
The authors provide the workflow, configuration, and container on GitHub (https://github.com/tomar-neeri/mtdna-heteroplasmy-host-reads) and archived the project on Zenodo (https://doi.org/10.5281/zenodo.23034211). Heteroplasmic variant calls and summary statistics are provided as Supplementary Data, and host-depleted metagenomic reads are available under BioProject PRJNA1478410.
The study reports ethics approval from the Institutional Ethics Committee for Research on Human Subjects (IEC) CSIR-NEERI Nagpur and states that participant consent and reporting guidelines were followed. The authors declared no competing interests. The preprint status is noted: this work has not been peer reviewed and should not guide clinical practice.