Table 1:
Leave-one-out validation for the TranNet Model.
| Cancer | BRCA | LUAD | LUSC | PRAD | LIHC |
|---|---|---|---|---|---|
|
| |||||
| Number/Percentage of predictors (optimized set) | 279/3% | 347/4% | 398/4% | 300/4% | 127/2% |
| Paired-sample T-statistic (optimized set) | 11.7 | 11.1 | 8.3 | 8.7 | 9.0 |
| Paired-sample P-value (optimized set) | 1.1e-20 | 8.7e-16 | 5.6e-11 | 2.6e-11 | 4.5e-11 |
|
| |||||
| The number of predictors (top 30%) | 2792 | 2607 | 2989 | 1919 | 2254 |
| Paired-sample T-statistic (top 30%) | 8.8 | 7.5 | 6.1 | 6.0 | 6.5 |
| Paired-sample P-value (top 30%) | 2.6e-14 | 4.9e-10 | 1.6e-07 | 2.7e-07 | 8.1e-08 |
The paired-sample T-test was used to compare the sample-wise prediction errors on real and permutated response data. This test was performed for the different numbers of top RP scoring predictors. The selected predictors corresponding to the optimal cut-off are provided with their RP scores in Additional file 6. The bottom rows report result using top 30% of predictors. The top rows report the results for the cancer-specific optimised set of predictors (See Figure 3A and Figure S1 in Additional file 1).