Table 1.
Metrics for identifying different types of mosaic variants from whole exome sequencing data. Values are based on ultra-high-depth sequencing validation results, and any uninformative results were excluded
| Type of mosaicism | Variant caller | Mosaic prioritisation | Validated mosaic | Validated constitutive or FP | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|---|---|---|
| De novo (child-PZM) | DNG | Binomial p < 0.0001 | 56 | 17 | 69% | 84% | 73% | 80% |
| Binomial p ≥ 0.0001 | 24 | 81 | ||||||
| VAF < 0.27 | 63 | 32 | 78% | 68% | 64% | 84% | ||
| VAF ≥ 0.27 | 17 | 66 | ||||||
| Inherited (low-level parent-PZM) | DNG | <2 alt reads in one parent | 3 | 57 | 63% | 89% | 42% | 95% |
| ≥2 alt reads in one parent | 5 | 7 | ||||||
| Inherited (high-level parent-PZM) | GATK | Binomial p < 0.0001 | 9 | 7 | 69% | 67% | 56% | 78% |
| Binomial p ≥ 0.0001 | 4 | 14 | ||||||
| VAF < 0.27 | 13 | 12 | 100% | 43% | 52% | 100% | ||
| VAF ≥ 0.27 | 0 | 9 |
PZM postzygotic mosaicism, DNG DeNovoGear, GATK Genome Analysis Toolkit, VAF variant allele fraction from exome sequencing data, FP false positive, Binomial p = binomial test on the alternative allele reads, centred around 0.5, PPV positive predictive value, NPV negative predictive value