To choose the number of clusters used by MCCA and LRACluster, we mistakenly took the solution with minimal silhouette score, rather than maximal.
Clustering with the maximal silhouette reduced the performance of both methods. We report here the results of the runs with maximal silhouette score.
-log10 log-rank p-value
MCCA has significant difference in survival in 4 of the 10 datasets.
LRACluster has significant difference in survival in 4 of the 10 datasets.
| Method Name |
AML |
BIC |
COAD |
GBM |
KIRC |
LIHC |
LUSC |
SKCM |
OV |
SARC |
Means |
| MCCA |
3.49 |
1.02 |
0.16 |
2.30 |
1.82 |
0.15 |
0.47 |
4.07 |
0.55 |
1.08 |
1.51 |
| LRACluster |
0.68 |
1.38 |
0.22 |
0.12 |
2.04 |
0.72 |
0.52 |
4.08 |
0.05 |
1.42 |
1.12 |
Number of enriched clinical parameters
MCCA has at least one enriched clinical parameter in 8 of the 10 datasets.
LRACluster has at least one enriched clinical parameter in 7 of the 10 datasets.
| Method Name |
AML |
BIC |
COAD |
GBM |
KIRC |
LIHC |
LUSC |
SKCM |
OV |
SARC |
Means |
| MCCA |
1 |
1 |
1 |
1 |
1 |
1 |
0 |
1 |
0 |
2 |
0.9 |
| LRACluster |
1 |
1 |
0 |
1 |
2 |
1 |
0 |
3 |
0 |
2 |
1.1 |
Number of clusters
| Method Name |
AML |
BIC |
COAD |
GBM |
KIRC |
LIHC |
LUSC |
SKCM |
OV |
SARC |
Means |
| MCCA |
3 |
2 |
7 |
2 |
3 |
2 |
2 |
6 |
2 |
2 |
3.1 |
| LRACluster |
3 |
2 |
2 |
2 |
9 |
2 |
2 |
2 |
2 |
3 |
2.9 |
