Table 3 Ablation evaluation results for different components of DMFF-DTA on the Davis dataset

From: Dual modality feature fused neural network integrating binding site information for drug target affinity prediction

 

MSE

CI

\({r}_{m}^{2}\)

W/o GEM

0.224

0.889

0.701

W/o LinkAttention

0.234

0.877

0.699

W/o Virtual node

0.230

0.886

0.695

W/o Source feature

0.222

0.891

0.699

W/o Warm up

0.231

0.885

0.696

W/o MEFseq

0.387

0.797

0.478

W/o MEFstr

0.233

0.880

0.685

Full model

0.218

0.894

0.702

  1. “W/o” means removing the corresponding component from the full model. The Full model is proposed DMFF-DTA. Bold indicates the best performance, and underline indicates the second best for each metric.
Back to article page