How to interpret VCF quality scores?

[idonaldson]

In [variant call format (VCF)][1] files produced at the end of the samtools mpileup variant detection [pipeline][2] there are two quality scores:

1) QUAL (col 6) = Phred based score that the variant shown in the ALT col is wrong.

2) INFO (col 8) MQ flag = RMA mapping quality

The two scores do not have a linear relationship. Some variants have a high mapping quality, but lower QUAL....

Can anyone describe how these two scores are created and how they are related?

I want to be able to filter variants based on these two scores, but do not fully understand what they mean. Should i filter based on both scores or just 'QUAL' for variants?

Thanks.


[1]: http://www.1000genomes.org/wiki/Anal...mat-version-41
[2]: http://samtools.sourceforge.net/mpileup.shtml

[swbarnes2]

I've been wondering that myself, and we've taken a stab at doing some sanger sequencing to see what quality values are associated with getting positive sanger confirmation.

But you are comparing apples and oranges here, I think. Mapping quality is about how likely it is that the read has mapped to the right place, and this has as much to do with your underlying genome as it does with the individual letter quality of the reads. So they shouldn't correlate much.

There is, however, another quality score, the GQ. My ad hoc observations suggest that the GQ is less sensitive to depth of coverage than the quality score, and this makes it less useful for determine which SNPs are real than the QUAL score.