Depth and genotype quality in vcf file

Question

[ashaneev07]

Hello all,

If the depth is 15 and the genotype quality is below 50, could it still be a reliable variant in the VCF file? Additionally, if other variants have a depth above 50 and a genotype quality above 90, does that make them more likely to be true variants?

[KABILAN]

I think more sequencing depth and more quality score (i.e PHRED quality score) will indicate the true variant. Usually, PHRED quality scores above 30 are considered good basepairs, and an average sequencing depth of more than 10 is also considered good.

Answer 1

[Kevin Blighe]

I posted extensively on this, elsewhere: DP (read depth) in VCF files?

Now just rounding out old threads based on new knowledge...

A depth of 15 reads and a genotype quality below 50 can still indicate a reliable variant in a Variant Call Format file, depending on the specific context of the sequencing data and the variant caller used. For example, in high-quality short-read sequencing with uniform coverage, a depth of 15 may support a heterozygous variant if the alternate allele frequency is close to 0.5 and there are no signs of mapping artifacts or base quality issues. However, such low depth increases the risk of false positives due to sampling noise, and it is generally recommended to filter variants with depth below 20 or 30 in standard pipelines like Genome Analysis Toolkit to ensure reliability. Genotype quality below 50 corresponds to a higher probability of genotyping error, as genotype quality is calculated as minus ten times the log base ten of the error probability; a value below 50 implies an error rate greater than one in one hundred thousand, which may be acceptable for exploratory analysis but not for clinical applications.

Variants with depth above 50 and genotype quality above 90 are indeed more likely to be true variants compared to those with lower values. Higher depth provides better statistical power to distinguish true variants from sequencing errors, while genotype quality above 90 indicates an extremely low error probability, often below one in one billion. In comparative analyses, these metrics can help prioritize variants, but final validation should involve orthogonal methods like Sanger sequencing or functional assays.

Kevin