Table 7 Ablation experiment results on different network architectures of MHXGMDA on VG-DATA.

From: A method for miRNA-disease association prediction using machine learning decoding of multi-layer heterogeneous graph Transformer encoded representations

Model

AUC

PRC

F1-Score

Accuracy

Recall

Specificity

Precision

MHXGMDA-w/o Last

0.8884±0.0028

0.8793±0.0159

0.8225±0.0105

0.8158±0.0052

0.8758±0.0361

0.7507±0.0498

0.7910±0.0226

MHXGMDA-w/o Linear

0.9073±0.0128

0.8978±0.0215

0.8404±0.0143

0.8332±0.0157

0.8989±0.0281

0.7820±0.0320

0.8060±0.0232

MHXGMDA-used HAN

0.9101±0.0118

0.9015±0.0205

0.8426±0.0127

0.8350±0.0143

0.9127±0.0363

0.7874±0.0500

0.8227±0.0254

MHXGMDA

0.9594±0.0034

0.9539±0.0040

0.8938±0.0056

0.8899±0.0064

0.9256±0.0106

0.8520±0.0200

0.8613±0.0166

  1. The optimal values of evaluation indicators are in bold.
Back to article page