Table 4 Results of the proposed model and baseline models with 300 and 500 genes for molecular subtype classification using the TCGA pan-cancer dataset
From: A multimodal graph neural network framework for cancer molecular subtype classification
Model | 300 | 500 | ||
|---|---|---|---|---|
Accu.a | F1 | Accu. | F1 | |
Proposed w/ GAT | 77.6% ± 1.6% | 0.76 ± 0.02 | 81.6% ± 1.2% | 0.80 ± 0.01 |
Proposed w/ GCN | 75.8% ± 1.1% | 0.74 ± 0.02 | 80.0% ± 1.2% | 0.79 ± 0.02 |
FC-NN | 65.9% ± 1.3% | 0.59 ± 0.04 | 77.5% ± 1.4% | 0.74 ± 0.02 |
GCN (Original) | 74.5% ± 1.6% | 0.72 ± 0.05 | 76.1% ± 1.3% | 0.73 ± 0.03 |
GCN (Modified) | 75.5% ± 1.4% | 0.72 ± 0.03 | 77.9% ± 1.1% | 0.77 ± 0.02 |
Multi-omics GCN (Original) | 76.4% ± 1.3% | 0.76 ± 0.03 | 77.4% ± 1.3% | 0.77 ± 0.03 |
Multi-omics GCN (Modified) | 77.4% ± 1.3% | 0.76 ± 0.02 | 78.2% ± 1.2% | 0.75 ± 0.02 |
GrAMME (Modified) | 77.4% ± 1.5% | 0.76 ± 0.02 | 79.6% ± 1.4% | 0.79 ± 0.02 |
Multi-omics GAT (Original) | 73.4% ± 1.8% | 0.71 ± 0.04 | 75.1% ± 1.5% | 0.74 ± 0.04 |
Multi-omics GAT (Modified) | 75.8% ± 1.5% | 0.74 ± 0.04 | 77.4% ± 1.3% | 0.74 ± 0.02 |