Skip to main content

Table 2 MAE performance of each model on the four tasks

From: A multitask GNN-based interpretable model for discovery of selective JAK inhibitors

 

Methods

Training set

Validation set

Test set

Global

MTATFP

0.15

0.37

0.37

STATFP

0.21

0.40

0.40

LightGBM_MD

0.22

0.45

0.44

LightGBM_ECFP4

0.23

0.43

0.43

JAK1

MTATFP

0.19

0.37

0.39

STATFP

0.18

0.36

0.39

LightGBM_MD

0.25

0.45

0.44

LightGBM_ECFP4

0.24

0.42

0.44

JAK2

MTATFP

0.18

0.38

0.38

STATFP

0.21

0.40

0.40

LightGBM_MD

0.22

0.48

0.47

LightGBM_ECFP4

0.25

0.46

0.46

JAK3

MTATFP

0.16

0.42

0.41

STATFP

0.27

0.49

0.47

LightGBM_MD

0.24

0.49

0.48

LightGBM_ECFP4

0.26

0.48

0.45

TYK2

MTATFP

0.15

0.30

0.30

STATFP

0.19

0.34

0.35

LightGBM_MD

0.15

0.36

0.38

LightGBM_ECFP4

0.18

0.37

0.37

  1. The lower the MAE value was, the better the model performed