Parametrization | Method | TN | FP | Sn |
---|
 | 1st-ATG | 688 | 15,433 | 0.0427 |
 | TISHunter | 0 | 16,121 | 0.0000 |
 | ATGpr | 0 | 16,121 | 0.0000 |
 | TIS Miner | 3,142 | 12,979 | 0.1949 |
 | NetStart | 575 | 15,546 | 0.0357 |
homogeneous | LLKR | 5,179 | 10,942 | 0.3213 |
 | WLLKR | 7,260 | 8,861 | 0.4503 |
 | MFCWLLKR | 1,688 | 14,433 | 0.1047 |
 | PFCWLLKR | 6,785 | 9,336 | 0.4209 |
 | BAYES | 6,813 | 9,308 | 0.4226 |
cluster-specific | LLKR | 6,385 | 9,736 | 0.3961 |
 | WLLKR | 8,080 | 8,041 | 0.5012 |
 | MFCWLLKR | 1,995 | 14,126 | 0.1238 |
 | PFCWLLKR | 8,685 | 7,436 | 0.5387 |
 | BAYES | 8,057 | 8,064 | 0.4998 |
random split | LLKR | 5,155 | 10,966 | 0.3198 |
 | WLLKR | 7,176 | 8,945 | 0.4451 |
 | MFCWLLKR | 1,687 | 14,434 | 0.1046 |
 | PFCWLLKR | 6,748 | 9,373 | 0.4186 |
 | BAYES | 6,824 | 9,297 | 0.4233 |
- 16,121 non-TIS-containing instances were used in three separate five-fold cross-validation experiments. Results are shown from applying a non-stratified parameter set (homogeneous), a priori-known cluster-specific parameter sets for k = 3 (cluster-specific), and group-specific parameter sets for a random three-way split of the data (random split). TN represents the number of instances for which the method (correctly) refused to predict a TIS, and FP denotes the number for which some prediction was made, though always incorrect (see Figure 2). .