Table 2.
Jaccard overlap between the selected pathways across the 4 study and the 6 study pathways LIMMA was used to normalize the data for subject identifier.
| Method | Time bin | Jaccard overlap | 4 studies | 6 studies |
|---|---|---|---|---|
| CPE | 1–8 | 0.0877 | 81 | 167 |
| LPE | 1–8 | 0.0588 | 60 | 138 |
| CEPA | 1–8 | 0.0169 | 60 | 60 |
| ORA | 1–8 | 0.2095 | 254 | 635 |
| CPE | 9–16 | 0.1014 | 75 | 153 |
| LPE | 9–16 | 0.0885 | 81 | 128 |
| CEPA | 9–16 | 0.0169 | 60 | 60 |
| ORA | 9–16 | 0.2288 | 227 | 507 |
| CPE | 17–24 | 0.1408 | 113 | 203 |
| LPE | 17–24 | 0.1186 | 89 | 175 |
| CEPA | 17–24 | 0.0435 | 60 | 60 |
| ORA | 17–24 | 0.3574 | 363 | 552 |
| CPE | 25–32 | 0.0846 | 74 | 144 |
| LPE | 25–32 | 0.1196 | 67 | 139 |
| CEPA | 25–32 | 0.0256 | 60 | 60 |
| ORA | 25–32 | 0.2222 | 235 | 502 |
The final two columns are the number of pathways selected by the method using the stated training dataset (4 or 6 studies). The CPE configuration is pre-computed, directed edges with PageRank centrality.