Skip to main content

Table 6 Comparison of all baseline approaches and ELECTRA-DTA on the KIBA datasets

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 KIBA Dataset KronRLS 0.782 0.411 - 0.342 0.635
SimBoost 0.836 0.222 - 0.629 0.760
DeepDTA 0.863 0.194 0.848 0.673 0.788
WideDTA 0.875 0.179 - - -
DeepCDA 0.889 (0.002) 0.176 0.855 0.682 (0.008) 0.812 (0.005)
ELECTRA-DTA 0.889 (0.003) 0.162 0.879 0.727 (0.004) 0.795 (0.006)
refined KIBA Dataset DeepDTA 0.892 (0.026) 0.152 0.896 0.766 (0.085) 0.798 (0.063)
Attention-DTA 0.880 (0.001) 0.158 0.883 0.742 (0.015) 0.795 (0.003)
ELECTRA-DTA 0.892 (0.002) 0.143 0.892 0.780 (0.014) 0.805 (0.005)
  1. Bold values represent the best performance over all competitive methods