Sample quality control¶
For a small proportion of samples, the sequencing data are not of sufficiently high quality to make reliable CNV calls.
Case flags¶
If CNV data for a proband do not pass this quality control step, the family is flagged in the Interpretation Portal with one of the following flags:
| Flag | Summary |
|---|---|
poor quality cnv calls |
majority of the CNV calls in the sample are expected to be of poor quality |
suspected poor quality cnv calls |
some of the CNV calls in the sample are expected to be of poor quality |
A poor quality cnv calls flag will take precedence over a suspected poor quality cnv calls flag.
Quality control metrics¶
Sample level quality control is performed based on the number and ratio of different call types and the proportion of common CNVs detected. The four metrics that we use to assess sample quality are:
- The count of autosomal PASS CNVs
- The log2 proportion of autosomal PASS LOSS CNVs to GAIN CNVs
- The proportion of autosomal PASS CNVs that are common, as defined by having a 50% reciprocal overlap with a CNV from Conrad et al. 2010.
- The coverage uniformity metric, as output by the DRAGEN software. Coverage uniformity is a measure of the evenness of coverage of a sample and ranges between a score of 0 and 1, with higher numbers indicating a less uniform sample.
Quality control metrics thresholds¶
The figures below outline the thresholds applied for each metric. The density curve represents the distribution of samples relative to these thresholds.
The case will have a poor quality cnv calls flag if the proband sample has:
- count of autosomal PASS CNVs ≥ 980 or
- Log2(Loss/Gain) < -2.2 or
- coverage uniformity metric ≥ 0.55
The case will have a suspected poor quality cnv calls flag if the proband sample has:
- count of autosomal PASS CNVs ≥ 195 and < 980, or ≤ 121 or
- Log2(Loss/Gain) ≥ -2.2 and ≤ -1.0, or ≥ 0.1 or
- fraction of common autosomal PASS CNV calls < 0.375
