Table 1 The table reports the root mean squared error (RMSE) on QM9, the mean absolute error (MAE) for ZINC, and R2 for DOCKSTRING (DOCK) and MoleculeNet, presented as mean ± standard deviation over 5 runs

From: An end-to-end attention-based approach for learning on graphs

 

Target

DropGIN

GAT

GATv2

PNA

Graphormer

TokenGT

GPS

ESA

QM9 ()

μ

0.55 ± 0.01

0.55 ± 0.01

0.55 ± 0.01

0.53 ± 0.01

0.63 ± 0.01

0.76 ± 0.02

0.94 ±  0.17

0.56 ± 0.00

 

α

0.44 ± 0.03

0.48 ± 0.02

0.46 ± 0.02

0.48 ± 0.07

0.40 ± 0.02

0.45 ± 0.01

0.60 ± 0.13

0.40 ± 0.00

 

ϵHOMO

0.11 ± 0.00

0.11 ± 0.00

0.10 ± 0.00

0.10 ± 0.00

0.11 ± 0.00

0.12 ± 0.00

0.11 ± 0.00

0.10 ± 0.00

 

ϵLUMO

0.12 ± 0.00

0.12 ± 0.00

0.13 ± 0.00

0.11 ± 0.00

0.11 ± 0.00

0.13 ± 0.00

0.11 ± 0.00

0.11 ± 0.00

 

Δϵ

0.17 ± 0.00

0.16 ± 0.01

0.16 ± 0.00

0.14 ± 0.00

0.16 ± 0.00

0.18 ± 0.01

0.15 ± 0.01

0.15 ± 0.00

 

R2

29.87 ± 0.34

30.91 ±  0.15

30.15 ± 0.36

28.50 ± 0.44

29.63 ± 0.35

31.54 ± 0.42

30.42 ±  1.19

28.33 ± 0.32

 

ZPVE

0.03 ± 0.00

0.03 ± 0.00

0.03 ± 0.00

0.03 ± 0.00

0.06 ± 0.03

0.03 ± 0.00

0.03 ± 0.01

0.03  ± 0.00

 

U0

29.82 ± 6.32

27.82 ± 3.98

25.40 ±  3.62

31.64 ±  6.49

24.60 ± 4.70

15.48 ±  4.67

14.50 ±  4.34

4.78 ± 0.67

 

U

24.74 ± 6.31

30.91 ±  2.62

24.92 ± 4.18

25.04 ±  5.48

15.55 ± 8.23

12.45 ± 1.86

15.82 ± 4.26

6.80 ± 2.81

 

H

34.02 ± 5.66

28.92 ±  4.11

23.42 ± 2.77

27.34 ±  8.17

19.01 ± 4.65

11.42 ± 3.72

12.92 ±  4.29

6.02  ± 1.32

 

G

25.41 ± 4.86

31.49 ±  1.86

23.24 ± 2.54

22.31 ±  3.45

31.20 ± 4.61

29.72 ± 8.79

13.08 ± 4.81

6.10  ± 1.33

 

cV

0.19 ± 0.01

0.19 ± 0.01

0.19 ± 0.00

0.18 ± 0.01

0.18 ± 0.02

0.18 ± 0.00

0.19 ± 0.03

0.16 ± 0.00

 

\({U}_{0}^{{{\rm{ATOM}}}}\)

0.38 ± 0.02

0.39 ± 0.03

0.38 ± 0.02

0.35 ± 0.02

0.29 ± 0.00

0.30 ± 0.02

0.33 ± 0.05

0.24 ± 0.01

 

UATOM

0.40 ± 0.04

0.48 ± 0.07

0.40 ± 0.04

0.36 ± 0.02

0.30 ± 0.01

0.30 ± 0.00

0.33 ± 0.06

0.24 ± 0.01

 

HATOM

0.38 ± 0.04

0.38 ± 0.02

0.41 ± 0.06

0.37 ± 0.03

0.30 ± 0.01

0.31 ± 0.01

0.36 ± 0.09

0.25 ± 0.00

 

GATOM

0.37 ± 0.04

0.34 ± 0.02

0.33 ± 0.01

0.31 ± 0.02

0.26 ± 0.01

0.27 ± 0.01

0.36 ± 0.08

0.22 ± 0.02

 

A

0.90 ± 0.06

0.97 ± 0.12

1.08 ± 0.16

1.01 ± 0.10

64.88 ±  29.77

3.82 ± 2.05

1.42 ± 0.44

0.75 ± 0.11

 

B

0.19 ± 0.05

0.21 ± 0.04

0.26 ± 0.01

0.26 ± 0.02

0.10 ± 0.03

0.11 ± 0.01

0.16 ± 0.05

0.08 ± 0.01

 

C

0.27 ± 0.02

0.27 ± 0.01

0.28 ± 0.00

0.28 ± 0.01

0.12 ± 0.04

0.10 ± 0.02

0.12 ± 0.05

0.05 ± 0.01

DOCK ()

ESR2

0.68 ± 0.00

0.67 ± 0.00

0.65 ± 0.00

0.70 ± 0.00

OOM

0.64 ± 0.01

0.68 ± 0.00

0.70 ± 0.00

 

F2

0.89 ± 0.00

0.89 ± 0.00

0.89 ± 0.00

0.89 ± 0.00

OOM

0.87 ± 0.01

0.88 ± 0.00

0.89 ± 0.00

 

KIT

0.83 ± 0.00

0.83 ± 0.00

0.83 ± 0.00

0.84 ± 0.00

OOM

0.80 ± 0.01

0.83 ± 0.00

0.84 ±  0.00

 

PARP1

0.92 ± 0.00

0.92 ± 0.00

0.92 ± 0.00

0.92 ± 0.00

OOM

0.91 ± 0.00

0.92 ± 0.01

0.93 ± 0.00

 

PGR

0.70 ± 0.00

0.68 ± 0.01

0.67 ± 0.01

0.72 ± 0.00

OOM

0.68 ± 0.01

0.70 ± 0.01

0.73 ± 0.00

MoleculeNet ()

FREESOLV

0.97 ± 0.00

0.96 ± 0.01

0.97 ± 0.01

0.95 ± 0.01

0.93 ± 0.00

0.93 ± 0.02

0.86 ± 0.03

0.98 ±  0.00

 

LIPO

0.81 ± 0.01

0.82 ± 0.01

0.82 ± 0.01

0.83 ± 0.01

0.61 ± 0.04

0.55 ± 0.02

0.79 ± 0.00

0.81 ±  0.01

 

ESOL

0.94 ± 0.01

0.93 ± 0.01

0.93 ± 0.00

0.94 ± 0.01

0.91 ± 0.02

0.89 ± 0.03

0.91 ± 0.00

0.94 ± 0.00

PCQM4MV2 ()

N/A

N/A

N/A

N/A

N/A

N/A

N/A

0.0235

Dataset ()

GAT

GATv2

PNA

Graphormer

TokenGT

GPS

ESA

ESA (PE)

ZINC

0.078 ± 0.01

0.079 ± 0.00

0.057 ± 0.01

0.036 ± 0.00

0.047 ± 0.01

0.024 ± 0.01

0.027 ± 0.00

0.015 ± 0.00

  1. The MAE is reported for PCQM4MV2 (it is the standard metric for this task) over a single run due to the size of the dataset. The lowest MAEs and RMSEs, the highest R2 values, and table headers are highlighted in bold. OOM denotes out-of-memory errors. A complete table, including GCN and GIN, is provided in Supplementary Table 1.
  2. Mean absolute error (MAE) on the full ZINC dataset, including an ESA model augmented with positional encodings (PE). A complete table, including GCN, GIN, and DropGIN is provided in Supplementary Table 3.