Table 2.
The performance comparison of GraphCpG with other methods on different datasets.a
| Dataset | Cell number | AUROC |
MCC score |
||||||
|---|---|---|---|---|---|---|---|---|---|
| Deep learning |
Deep learning |
||||||||
| CaMelia | DeepCpG | CpG Transformer | GraphCpG | CaMelia | DeepCpG | CpG Transformer | GraphCpG | ||
| HCC | 25 | 97.11 | 96.01 | 97.56 | 96.99 | 83.32 | 78.71 | 84.43 | 81.61 |
| MBL | 30 | 89.36 | 87.12 | 92.05 | 89.73 | 63.17 | 60.09 | 70.58 | 64.71 |
| Hemato | 122 | 87.68 | 88.26 | 89.56 | 89.77 | 69.04 | 67.96 | 68.15 | 69.05 |
| Neuron-Mouse | 690 | 91.13 | 88.59 | 90.87 | 91.75 | 71.05 | 66.52 | 70.77 | 71.1 |
| Neuron-Homo | 780 | 92.98 | 90.06 | 92.31 | 93.2 | 75.01 | 73.85 | 75.15 | 75.24 |
Bold numbers indicate the best performance.