Table 4.
Prediction results in Mallick data, when using half of the sample size in the PRISM study as the training set for 10 times and the data in NLIBD study as the testing set.
| Mallick cohort (NM = 466) PRISM (training, n = 77 or 78) NLIBD (testing, n = 65) | Training (ENVIM) | Training (MelonnPan) | Testing (ENVIM) | Testing (MelonnPan) |
|---|---|---|---|---|
| Seed1 | 429 (92%) | 161 (35%) | 147 (32%) | 96 (21%) |
| Seed2 | 402 (86%) | 202 (43%) | 162 (35%) | 104 (22%) |
| Seed3 | 427 (92%) | 164 (35%) | 140 (30%) | 92 (20%) |
| Seed4 | 428 (92%) | 199 (43%) | 148 (32%) | 97 (21%) |
| Seed5 | 439 (94%) | 211 (45%) | 160 (34%) | 111 (24%) |
| Seed6 | 427 (92%) | 180 (39%) | 157 (34%) | 113 (24%) |
| Seed7 | 424 (91%) | 178 (38%) | 143 (31%) | 98 (21%) |
| Seed8 | 424 (91%) | 150 (32%) | 142 (30%) | 98 (21%) |
| Seed9 | 425 (91%) | 159 (34%) | 150 (32%) | 101 (22%) |
| Seed10 | 419 (90%) | 181 (39%) | 152 (33%) | 105 (23%) |
| Mean | 424 (91%) | 179 (38%) | 150 (32%) | 102 (22%) |