Table 3. Sarek WGS somatic variant benchmarking.
Summary of accuracy measures for the two somatic variant callers used in Sarek to detect somatic single-base mutations (SSMs) and somatic insertion/deletion mutations (SIMs), as well as their union and intersection.
| Somatic caller | Recall | Precision | F1-score |
|---|---|---|---|
| SSM (Gold Set: n=1263) | |||
| GATK4 Mutect2 | 0.80 | 0.45 | 0.58 |
| Strelka2 | 0.77 | 0.29 | 0.42 |
| Union (GATK4 Mutect2, Strelka2) | 0.82 | 0.23 | 0.36 |
| Intersection (GATK4 Mutect2, Strelka2) | 0.74 | 0.88 | 0.80 |
| Benchmark median * | 0.68 | 0.78 | 0.71 |
| SIM (Gold Set: n=347) | |||
| GATK4 Mutect2 | 0.48 | 0.38 | 0.42 |
| Strelka2 | 0.74 | 0.31 | 0.44 |
| Union (GATK4 Mutect2, Strelka2) | 0.77 | 0.25 | 0.38 |
| Intersection (GATK4 Mutect2, Strelka2) | 0.46 | 0.77 | 0.58 |
| Benchmark median * | 0.34 | 0.71 | 0.48 |
* The median accuracy measures across 18 somatic variant calling procedures as previously reported ( Alioto et al., 2015)