Inquiry on polyploid support and mosaicism detection

[HLHsieh]

Hi,

Thank you for developing this valuable tool. I have a few questions regarding its capabilities:
1. Does TRGT support polyploid genomes? If so, are there specific parameters or settings recommended for analyzing such data?
2. Can TRGT be used to identify or evaluate mosaic repeat expansions? Specifically, are there any metrics, visual outputs, or recommended practices to help determine whether a repeat expansion is mosaic rather than due to technical noise?
3. Does TRGT provide per-read repeat counts or other read-level outputs that would allow further evaluation?

I would greatly appreciate any insights or suggestions on how to best apply TRGT in these scenarios.

Thank you for your time and assistance.

Best regards,
Hsin

[pbsena]

Hello,

Thanks for the questions.

1. Currently TRGT does have the underlying assumption that there are at most two alleles, and it honors the karyotype (XX/XY) passed by the user to genotype x-chromosome loci
2. Currently we do not have an exact definition of mosaicism, but the output data allows you do to various quantifications that pertain to mosaicism. (1) The ALLR tag in the VCF shows the length range of repeats assigned to each allele (2) the .spanning.bam output has the TR and AL tag that allows you to stratify the repeat sequences by locus and allele, so, for example, to get the length (not motif count) of all reads that map to allele 1 in the FXN locus in output reads.spanning.bam you could do

samtools view reads.spanning.bam | grep "TR:Z:FXN" | grep "AL:i:1" | awk '{print length($10)}'

additionally, the trgt plot tools, especially with --plot-type=waterfall mode, allows you to visually assess the variation in TR length per sequence.

3. We currently do not report per-read motif counts (just for the consensus), but by stratifying reads like in (2) you can isolate the repeat sequences of each read and count motifs. We are working on creating more detailed summary statistics in the VCF file that represent variation in other ways (e.g., standard deviation of lengths, or 95% confidence intervals), but it is unlikely that VCFs will report per-read values as the format itself is ill-equipped to handle that amount of information.

Hope this helps, we appreciate the feedback as it's useful for us to know the user needs, I'd be happy to help with guidelines for downstream analyses using TRGT's outputs.