Table 3.
Accuracy of small indel calling, comparing assembly-based calling with two mapping-based approaches on the same libraries’ linked read data, one each from NA12878 (L3) and NA24385 (L5).
| Small indels | True positives | False negatives | False positives | Genotype Mismatch | Precision | Recall | F1 | |
|---|---|---|---|---|---|---|---|---|
| L3 (NA 12878) | Aquila | 499,301 | 32,081 | 40,292 | 9493 | 0.925 | 0.940 | 0.932 |
| FreeBayes | 419,344 | 80,354 | 45,977 | 39,636 | 0.903 | 0.839 | 0.870 | |
| Longranger | 463,732 | 35,966 | 88,431 | 22,513 | 0.843 | 0.928 | 0.883 | |
| L5 (NA 24385) | Aquila | 476,139 | 26,914 | 35,315 | 7986 | 0.931 | 0.946 | 0.939 |
| FreeBayes | 400,440 | 75,170 | 41,475 | 36,293 | 0.908 | 0.842 | 0.874 | |
| Longranger | 443,107 | 32,504 | 81,044 | 19,720 | 0.848 | 0.932 | 0.888 | |
| L5+L6 (NA 24385) | Aquila | 473,895 | 29,158 | 15,724 | 5335 | 0.968 | 0.942 | 0.955 |
The benchmark is GiaB v3.3.2 within the high-confidence regions. L5 + L6 can be achieved by Aquila through a multiple-library assembly mode, which is not applicable for other tools.