Table 3.
Descriptive statistics of the disease-causing variant ranking in the inherited retinal disease (IRD) patient whole-exome sequencing (WES) dataset for the four successfully run phenotype-aware variant prioritisation (VP) software tools and pairwise agreement with LIRICAL (reference)a
| VP software tool |
Patients with disease-causing variants filtered out/not prioritised |
Mean rank (SD) |
Median rank |
Min rank |
Max rank |
Top rank, % (n = 134) |
Agreement with LIRICALb (%) |
Disagreement with LIRICALb(%) |
Cohen's kappa (Agreement) |
Stuart-Maxwell test |
|---|---|---|---|---|---|---|---|---|---|---|
| LIRICAL | 4 | 1.7 (2.0) | 1 | 1 | 18 | 74.6 | Reference | |||
| Exomiser | 6 | 3.1 (11.7) | 1 | 1 | 123 | 73.9 | 68.7 | Better: 14.9 Worse: 16.4 |
0.25 (fair agreement) | 0.4189 |
| Xrare | 16 | 1.4 (0.8) | 1 | 1 | 5.5 | 73.1 | 61.9 | Better: 17.9 Worse: 20.1 |
0.10 (slight agreement) | 0.0016 |
| PhenIX | 6 | 1.8 (2.5) | 1 | 1 | 23 | 69.4 | 67.2 | Better: 14.2 Worse: 18.7 |
0.25 (fair agreement) | 0.3894 |
a
The mean/median/min/max ranks in the table refer to the effective n, that is, 134 minus the number of patients with disease-causing variants filtered out. ‘Top rank, %’ describes the percentage of disease-causing variants ranked as first out of the 134 exomes tested.
b
LIRICAL was chosen as the reference, being the best performing VP software tool at ranking the disease-causing variants overall.