Table 2.
Performances of Scuba, MKL1class and ProDiGe in the unbiased setting of Börnigen et al. [20]
Tool/Method | Rank | Rank | TPR in top | TPR in top | TPR in top | AUC | Rank difference |
---|---|---|---|---|---|---|---|
median | average | 5% (%) | 10% (%) | 30% (%) | p-value | ||
Genome-wide prioritization methods | |||||||
Scuba | 10.55 | 20.48 ± 23.53 | 33.3 | 47.6 | 78.6 | 0.80 | - |
MKL1class [12] | 13.30 | 23.42 ± 23.23 | 21.4 | 47.6 | 69.0 | 0.77 | 2.5 ·10−2 * |
ProDiGe [13] | 11.73 | 24.45 ± 27.33 | 31.0 | 45.2 | 71.4 | 0.76 | 3.0 ·10−7 * |
Candidate set-based prioritization methods | |||||||
Scuba | 12.95 | 23.32 ± 25.46 | 28.6 | 45.2 | 73.8 | 0.78 | - |
MKL1class [12] | 15.07 | 25.63 ± 24.73 | 23.8 | 40.5 | 61.9 | 0.76 | 9.7 ·10−2 |
ProDiGe [13] | 14.41 | 26.39 ± 29.09 | 26.2 | 40.5 | 71.4 | 0.75 | 2.7 ·10−3 * |
Values refer to predictions on all the 42 gene-disease associations. Rank difference p-values were obtained using Wilcoxon signed rank tests comparing separately Scuba/MKL1class and Scuba/ProDiGe ranks differences. Asterisks indicate significance of the tests at a threshold of 0.05
Italics indicates the top ranking score of each column