Skip to main content

Table 2 Predictions of 3,591 functionally validated single nucleotide variants by 15 mutation effect prediction algorithms

From: Benchmarking mutation effect prediction algorithms using functionally validated cancer-related missense mutations

Prediction algorithm

Prediction class

Functional categories

Total

Neutral

Non-neutral

Uncertain

(n = 140)

(n = 849)

(n = 2,602)

(n = 3,591)

CHASM (breast)

Driver

27

764

1,085

1,876

Passenger

113

85

1,517

1,715

CHASM (lung)

Driver

32

783

1,155

1,970

Passenger

108

66

1,447

1,621

CHASM (melanoma)

Driver

48

795

1,440

2,283

Passenger

92

54

1,162

1,308

FATHMM (cancer)

CANCER

71

831

1,655

2,557

PASSENGER/OTHER

69

18

947

1,034

FATHMM (missense)

Damaging

69

745

1,416

2,230

Tolerated

71

104

1,167

1,342

No weights

0

0

19

19

Mutation Assessor

High

2

71

97

170

Medium

50

579

1,053

1,682

Low

51

129

919

1,099

Neutral

37

69

527

633

N/A

0

1

6

7

MutationTaster

Disease_causing

34

740

1,313

2,087

Disease_causing_automatic

1

31

4

36

Polymorphism

99

78

1,285

1,462

Polymorphism_automatic

6

0

0

6

PolyPhen-2

Probably damaging

40

600

920

1,560

Possibly damaging

26

115

478

619

Benign

74

134

1,204

1,412

PROVEAN

Deleterious

43

632

955

1,630

Neutral

97

217

1,647

1,961

SIFT

Damaging

70

731

1,469

2,270

Tolerated

70

118

1,133

1,321

VEST

Functional

100

702

1,663

2,465

Neutral

40

147

939

1,126

CanDrA (breast)

Driver

140

805

2,423

3,368

Passenger

0

39

140

179

No-call

0

5

39

44

CanDrA (lung)

Driver

24

767

1,150

1,941

Passenger

102

59

1,282

1,443

No-call

14

23

170

207

CanDrA (melanoma)

Driver

28

734

1,147

1,909

Passenger

97

75

1,260

1,432

No-call

15

40

195

250

Condel

Deleterious

77

786

1,741

2,604

Neutral

63

63

861

987

  1. Single nucleotide variants (SNVs) were classified as non-neutral, neutral or uncertain based on functional/experimental data from the literature or mutation databases [28-30]. Each SNV was classified by each of the mutation effect prediction algorithms independently.