Table 2.
Average FOSCTTM of algorithms for integrating GE and DM data
| Cancer/Alg | INTEND | LIGER | Seurat v3 | MMD-MA | MMD-MA WFCI (500) | MMD-MA WFCI (2000) | JLMA WFCI (500) | JLMA WFCI (2000) |
|---|---|---|---|---|---|---|---|---|
| AML | 2.42 (25) | 29.83 (7) | 17.05 (36) | 23.63 (40) | 19.08 (40) | 22.35 (40) | 24.01 (8) | 28.38 (7) |
| BLCA | 0.04 (39) | 39.62 (9) | 13.86 (40) | 11.20 (40) | 16.34 (40) | 14.58 (40) | 34.80 (40) | 37.11 (40) |
| COAD | 0.02 (37) | 26.84 (19) | 19.14 (40) | 12.59 (40) | 12.19 (40) | 12.92 (40) | 32.98 (5) | 34.73 (4) |
| LGG | 6.82 (22) | 41.97 (8) | 32.06 (26) | 8.88 (40) | 15.50 (40) | 12.08 (40) | 37.41 (14) | 32.38 (12) |
| LIHC | 0.14 (36) | 42.34 (3) | 19.23 (38) | 16.04 (30) | 11.02 (30) | 12.94 (30) | 32.68 (21) | 36.03 (12) |
| LUAD | 0.06 (32) | 36.72 (4) | 16.36 (39) | 8.71 (40) | 14.11 (40) | 13.89 (40) | 29.60 (9) | 32.16 (8) |
| PAAD | 0.55 (30) | 36.68 (15) | 24.18 (35) | 11.07 (40) | 23.42 (40) | 16.27 (40) | 29.83 (3) | 27.44 (2) |
| PRAD | 0.37 (38) | 35.96 (8) | 16.32 (17) | 10.88 (40) | 11.15 (40) | 10.99 (40) | 27.14 (2) | 29.53 (2) |
| SARC | 0.05 (35) | 42.06 (15) | 12.86 (36) | 8.86 (40) | 20.97 (40) | 17.42 (40) | 34.47 (7) | 34.73 (5) |
| SKCM | 0.03 (39) | 42.20 (17) | 18.97 (37) | 16.02 (40) | 20.53 (40) | 16.62 (40) | 32.11 (15) | 34.71 (3) |
| THCA | 3.07 (11) | 32.58 (7) | 15.96 (36) | 6.71 (40) | 7.78 (40) | 6.65 (40) | 30.95 (2) | 27.52 (5) |
| Average (all datasets) | 1.23 (31) | 36.98 (10) | 18.73 (34) | 12.24 (39) | 15.64 (39) | 14.25 (39) | 31.45 (11) | 32.25 (9) |
Average FOSCTTM score (percentge) for each algorithm on each of the 11 cancer datasets. The optimal score is 0%, and the expected score for a random projection is 50%. The requested shared space dimension d ranges from 2 to 40 for each algorithm. The score shown is the best across all values of d, and the optimal d is given in parentheses. The numbers 500 and 2000 for MMD-MA and JLMA denote the number of selected genes in the WFCI runs.