Skip to main content

Advertisement

Table 1 Overview of the performance of the benchmarked methods expressed as z-scores per experiment

From: Beyond the hype: deep neural networks outperform established methods using a ChEMBL bioactivity benchmark set

Method MCC random BEDROC random MCC temporal BEDROC temporal Average SEM
NB 10 μM −2.41 −2.22 −2.07 −0.67 −1.84 0.40
NB −0.65 −0.66 −0.81 −0.64 −0.69 0.04
RF 0.56 −0.30 0.02 −1.41 −0.28 0.41
RF_PCM 0.88 −0.17 −0.46 −1.10 −0.21 0.41
SVM 0.11 0.36 0.53 0.30 0.32 0.09
LR 0.17 0.40 0.11 0.19 0.22 0.06
DNN 0.32 0.75 0.56 0.79 0.60 0.11
DNN_MC 0.60 0.85 1.03 1.20 0.92 0.13
DNN_PCM 0.44 0.98 1.09 1.33 0.96 0.19
  1. Z-scores are shown for all methods for both types of splitting and for both MCC and BEDROC. In italics the best performance for a given machine learning algorithm per column is highlighted. See main text for further details