Skip to content

Genome-wide copy number variant calls quality check

On rare occasions, genome sequencing data that satisfies the contractual coverage metrics can still have uneven coverage across the chromosomes (wavy or spiky in appearance). This suggests that CNV detection in that sample may be compromised. To identify this risk, sample level CNV quality is assessed based on the number and ratio of different types of CNVs called, as described below. Samples that fail any of the defined thresholds are flagged in VRT.

Samples are flagged with the poor quality CNV calls flag when most CNV calls are expected to be of poor quality. This flag is applied for the samples that have one or more of the following:

  • count of autosomal PASS CNVs ≥ 980
  • Log2(Loss/Gain) < -2.2

Samples are flagged with the “suspected poor quality CNV calls” flag if they may have an increased number of poor quality CNV calls. This flag is applied for samples that pass thresholds above, but have one or more of the following:

  • count of autosomal PASS CNVs ≥ 195 & < 980
  • count of autosomal PASS CNVs ≤ 121
  • Log2(Loss/Gain) ≤ -1.0 & >= -2.2
  • Log2(Loss/Gain) ≥ 0.1
  • the fraction of common autosomal PASS CNV calls is < 0.375. For this purpose, a CNV is defined as common if it has 50% reciprocal overlap with a CNV from Conrad et al. 2010.

The VRT user guide has further information on the displayed flags, and the Variant Review SOP instructs how clinical scientists should handle these.