What are the risque of keeping variants which read have low mapping quality?

Question

[lacb]

Hello,

I have a pipeline to do variant calling and a some point there is the GATK FilterMutectCalls tool which annotate the variants, and some get flagged "map_qual" when the mapping quality is low. By default, I remove them from the variant calling results. But for some variants I noticed there is a very high number of reads with good alignment quality, they "looks" good on IGV. So my question is:

What is the risk of keeping those variants in my variant calls? I'm not sure I understand well what does mapping quality represents.

Answer 1

[dthorbur]

Mapping quality refers to the confidence of the tool that the reads is correctly mapped to the reference sequence. There are a lot of reasons why this could have been flagged as low quality, but includes multi-mapping (i.e., when the read maps well to multiple different regions of the genome), low sequence PHRED scores, high mismatch rate, read length, etc...

You can introduce false positives and incorrectly place variants with multiple mapping locations. I'm sure there are more implications, but those are off the top of my head.

There are some cases where ignoring this flag can be useful, but I would suggest to proceed with caution. Highly variable or repetitive loci can be considered since mapping to these locations is difficult in the best of times, for example.