Table 3. Changes in network selections.
| Simulation Settings | Technique | Sensitivity | Specificity | # Selected | Filtering+SCCA Run time (min) | |
|---|---|---|---|---|---|---|
| Signal Strength | Total # Features | |||||
| Weak | 45305 | No Filter | 0.32 | 0.993 | x = 203, g = 103 | 31.7 |
| Weak | 45305 | SuMO-Fil | 0.80 | 0.996 | x = 118, g = 100 | 14.4 + 28.8 |
| Weak | 45305 | Low Means | 0.04 | 0.995 | x = 118, g = 94 | <0.1 + 35.7 |
| Weak | 45305 | Low Variance | 0.20 | 0.994 | x = 194, g = 95 | 0.3 + 39.2 |
| Moderate | 45305 | No Filter | 0.92 | 0.991 | x = 237, g = 215 | 30.9 |
| Moderate | 45305 | SuMO-Fil | 0.92 | 0.992 | x = 188, g = 190 | 15.1 + 31.1 |
| Moderate | 45305 | Low Means | 0.92 | 0.991 | x = 219, g = 199 | <0.1 + 33.9 |
| Moderate | 45305 | Low Variance | 0 | 0.989 | x = 293, g = 226 | 0.2 + 37.2 |
| Strong | 45305 | No Filter | 1 | 0.972 | x = 706, g = 576 | 32.3 |
| Strong | 45305 | SuMO-Fil | 1 | 0.980 | x = 499, g = 409 | 15.8 + 21.8 |
| Strong | 45305 | Low Means | 1 | 0.973 | x = 676, g = 551 | 0.1 + 38.7 |
| Strong | 45305 | Low Variance | 0 | 0.898 | x = 2576, g = 2036 | 0.2 + 40.6 |
| Weak | 25305 | No Filter | 0 | 0.996 | x = 72, g = 42 | 17.8 |
| Weak | 25305 | SuMO-Fil | 0 | 0.996 | x = 62, g = 50 | 4.3 + 13.4 |
| Weak | 25305 | Low Means | 0 | 0.995 | x = 70, g = 52 | <0.1 + 15.6 |
| Weak | 25305 | Low Variance | 0 | 0.995 | x = 72, g = 56 | <0.1 + 16.0 |
| Moderate | 25305 | No Filter | 1 | 0.993 | x = 114, g = 93 | 16.6 |
| Moderate | 25305 | SuMO-Fil | 0.96 | 0.994 | x = 96, g = 74 | 4.3 + 13.0 |
| Moderate | 25305 | Low Means | 0.96 | 0.994 | x = 107, g = 90 | <0.1 + 15.7 |
| Moderate | 25305 | Low Variance | 0 | 0.967 | x = 504, g = 336 | <0.1 + 16.3 |
| Strong | 25305 | No Filter | 1 | 0.977 | x = 356, g = 254 | 17.0 |
| Strong | 25305 | SuMO-Fil | 1 | 0.992 | x = 117, g = 99 | 4.3 + 9.2 |
| Strong | 25305 | Low Means | 1 | 0.978 | x = 329, g = 241 | <0.1 + 16.1 |
| Strong | 25305 | Low Variance | 0 | 0.979 | x = 315, g = 215 | <0.1 + 16.4 |
| Weak | 15305 | No Filter | 0 | 0.996 | x = 17, g = 19 | 6.5 |
| Weak | 15305 | SuMO-Fil | 0 | 0.997 | x = 14, g = 17 | 0.7 + 5.0 |
| Weak | 15305 | Low Means | 0 | 0.997 | x = 17, g = 18 | <0.1 + 6.1 |
| Weak | 15305 | Low Variance | 0 | 0.997 | x = 17, g = 18 | <0.1 + 6.0 |
| Moderate | 15305 | No Filter | 0.920 | 0.999 | x = 14, g = 23 | 6.6 |
| Moderate | 15305 | SuMO-Fil | 1 | 0.987 | x = 64, g = 78 | 0.7 + 4.5 |
| Moderate | 15305 | Low Means | 0.92 | 0.981 | x = 84, g = 107 | <0.1 + 6.0 |
| Moderate | 15305 | Low Variance | 0.04 | 0.954 | x = 242, g = 238 | <0.1 + 6.2 |
| Strong | 15305 | No Filter | 0 | 0.997 | x = 15, g = 19 | 6.7 |
| Strong | 15305 | SuMO-Fil | 0.720 | 1 | x = 10, g = 9 | 0.7 + 3.6 |
| Strong | 15305 | Low Means | 0 | 0.998 | x = 14, g = 12 | <0.1 + 6.2 |
| Strong | 15305 | Low Variance | 0 | 0.861 | x = 722, g = 709 | <0.1 + 6.0 |
The median network results after applying supervised SCCA [15] to 5 simulations under each simulation setting. Supervised SCCA was applied to the simulations before filtering and after filtering based on SuMO-Fil, low means, and low variance. Under the simulations with 45305 total features, SuMO-Fil at minimum maintained sensitivity (while improving sensitivity under weak signal strength) and increased specificity by reducing the amount of noise in the network selections. SuMO-Fil also maintained or increased specificity under simulations with 25305 total features, although sensitivity was decreased under both SuMO-Fil and low mean filters with moderate signal strength. SuMO-Fil maintained or increased sensitivity under simulations with 15305 total features, although specificity was decreased under moderate network signal. Filtering based on low variance often decreased sensitivity to 0 while drastically increasing the number of noise selected and increasing run times. Performing supervised SCCA after SuMO-Fil produced the shortest run times under most simulation settings.