Best Practices for Endosymbiont Variant Detection- Microbial vs Mitochondria?

[younghwankwak]

Hi GATK Community,

I'm working on a metagenomic dataset from an insect host and two obligate endosymbionts (non free-living bacteria with genome size < 200kb). These microbes are strictly vertically transmitted, with highly reduced genomes, and share some features with mitochondria — including clonal inheritance, high within-host coverage, and presumed low diversity.

To call intra-host variants, I applied both the mitochondrial and microbial pipelines with Mutect2 following GATK Best Practices. I noticed that two variant datasets seem substantially different (mito-mode has significantly more variants detected. But there are also some unique variants only found in micro-mode).

My specific questions are:

1. Which mode is more appropriate for detecting intra-host variation in endosymbionts like these? The metagenomic reads are sequenced from a single host

2. Does microbial mode introduce assumptions (e.g. error models, filter relaxation) that could inflate false positives?

3. Would you recommend combining both pipelines, or priotizing one?

I'm aiming to balance sensitivity and specificity. I'd really appreciate any input on choosing the most appropriate calling strategy for this system.