Skip to main content

Table 4 Performance of different methods on the independent S350 dataset.

From: NeEMO: a method using residue interaction networks to improve prediction of protein stability upon mutation

 

All mutations

Common mutations

Common mutations -10%

 

n

r

σ

n

r

σ

n

r

σ

Automute

315

0.46

1.42

299

0.44

1.45

264

0.60

1.06

CUPSAT

346

0.37

1.46

299

0.37

1.50

264

0.50

1.10

Dmutant

350

0.48

1.38

299

0.46

1.44

264

0.63

1.05

Eris

334

0.35

1.49

299

0.35

1.52

264

0.55

1.07

I-Mutant 2.0

346

0.29

1.50

299

0.27

1.56

264

0.39

1.16

I-Mutant 3.0

338

0.53

1.35

299

0.53

1.37

264

0.71

1.00

MuPro

350

0.41

1.43

299

0.41

1.48

264

0.49

1.12

PoPMuSiC 1.0

350

0.62

1.23

299

0.63

1.26

264

0.72

0.93

PoPMuSiC 2.0

350

0.67

1.16

299

0.67

1.21

264

0.80

0.86

NeEMO

350

0.67

1.16

299

0.68

1.19

264

0.79

0.88

  1. The comparison is reported (a) for all the mutations in the dataset, (b) the maximal subset of mutations where each tool is able to make a prediction and (c) the maximal subset where 10% of outliers are removed. The number of mutations (n) is shown together with the Pearson correlation (r) and distance from the real ΔΔG values (σ). The best prediction in each column is shown in bold.