ASK: How can I filter controversal imputed result?

[duaghk]

Hello,

I am currently using GLIMPSE2 for rare variant detection and came across a confusing result.

In the attached image, you can see that there are multiple imputed results (3 rows) for the same genomic position (chr1:10397). Each result appears to differ in the REF and ALT alleles, while the genotype calls for all three are 0|0.

Here are my questions:

How should I interpret these results? Why are there multiple imputation outcomes for the same position?
How should I filter or prioritize these results? For downstream analysis, which entry should I use or trust in this case?
If you have any suggestions on how to handle or clean such data, I’d greatly appreciate it.

I used GLIMPSE v2.0.0, and input vcf was generated by: LP-WGS -> DeepVariant call -> Filter PASS & DP >=5

Thank you in advance!

[srubinacci]

Hi,

Your reference panel contain a complex indel (STR maybe?) and the position is represented as 4 different bi-allelic indels and a SNP. This is one of the cases where you typically want to filter out those INDEL variants from the reference panel, as very likely quite unreliable at the position (and probably not well phased too, for the same reason).

Nothing wrong with GLIMPSE here. If unsure, always restrain your reference panel to INDELS to avoid difficult situations like this.
Simone