Skip to main content

Table 8 Comparison of all baseline approaches and the ELECTRA-DTA on the BindingDB Dataset

From: ELECTRA-DTA: a new compound-protein binding affinity prediction model based on the contextualized sequence encoding

Dataset Model CI MSE R \(r_m^2\) AUPR
Original BindingDB Dataset DeepDTA 0.812 (0.002) 0.832 0.824 0.623 (0.02) 0.443 (0.01)
DeepCDA 0.822 (0.001) 0.844 0.808 0.631 (0.002) 0.459 (0.003)
ELECTRA-DTA 0.832 (0.004) 0.693 0.852 0.645 (0.012) 0.807 (0.002)
refined BindingDB Dataset DeepDTA 0.826 (0.001) 0.703 0.845 0.669 (0.004) 0.795 (0.003)
Attention-DTA 0.804 (0.003) 0.844 0.811 0.619 (0.009) 0.764 (0.004)
ELECTRA-DTA 0.837 (0.002) 0.650 0.860 0.670 (0.027) 0.811 (0.001)
  1. Bold values represent the best performance over all competitive methods