Skip to main content
. 2023 Mar 23;14:1159076. doi: 10.3389/fmicb.2023.1159076

Table 1.

Comparison performance between our model and state-of-the-art models based on HMDAD dataset.

Methods AUC(5-fold cv) AUC(2-fold cv)
KATZHMDA (Zhu et al., 2021) (network-based) 0.8703±0.0199 0.8755±0.0103
NTSHMDA (Luo and Long, 2018) (network-based) 0.8982±0.0312 0.8615±0.0151
NGRHMDA (Huang et al., 2017) (binary local features-based) 0.8921±0.0327 0.8929±0.0059
BiRWMP (Luo and Xiao, 2017) (binary local features-based) 0.8777±0.0089 0.8698±0.0079
GRNMFHMDA (He et al., 2018) (matrix factorization-based) 0.8806±0.0156 0.8756±0.0164
GATMDA (Long et al., 2021) (graph neural network-based) 0.9554±0.0184 0.9538±0.0049
Our model 0.9867±0.0078 0.9708±0.0117