GangSTR gives very high GB values. Possibly related to piling FRRs

[gymreklab]

GangSTR gives very high GB values. Possibly related to piling FRRs

[nmmsv]

In genomewide more, off-target regions are not scanned. So that shouldn't change anything.
This regions seems a bit odd. It seems to have a massively inflated coverage, while it seems like there are a lot of mismatches in the aligned reads. While a large expansion could cause a fairly large inflation in coverage, it's also possible that this region is poorly assembled in the reference. We identified quite a few of these on hg19.
If you have other samples, you can compare the estimate on the same locus across samples. If it is called as expanded across your cohort, it's likely caused by an issue in the reference, and you can confidently filter this locus out.
Hope this helps! :)

[nmmsv]

Hi Matthias,
Thanks for reporting this issue.
Do you have off-target regions specified in your bed file?
Regardless of that, if you use --genomewide, the off-target regions should not be scanned.
I advise checking the loci that seem to be problematic in IGV. Sometimes the flanking region around a STR locus looks similar to imperfect copies of the motif. This can cause inflation in the estimate for that region. In that case, the best approach is to not included those loci, as GangSTR requires the flanking region to be non-repetitive.
Hope this helps.
Best,
Nima

[Matthlas]

Hi Nima,
Thanks for your answer.
I did not specify any off-target regions in my bed file yet. If that might help I would like to try that though. Do you have a recommendation on how to go about that? Would you recommend specifying all regions with the same motif for each locus or filtering them by read depth or something similar and add a subset as off-target regions to all other loci?
I already checked some of the loci in IGV and the flanking regions did not look particularly similar to the repeat.
Eg:

Thanks,
Matthias