Skip to main content

Table 2 Dependency on the R a value

From: The power metric: a new statistically robust enrichment-type metric for virtual screening applications with early recovery capability

Metric

R a

χ (%)

λ

0.01 (n = 50; N = 5000)

0.05 (n = 250; N = 5000)

0.2 (n = 1000; N = 5000)

PM

0.39 ± 0.36

0.57 ± 0.15

0.62 ± 0.07

1

1

ROCE

1.59 ± 1.83

1.62 ± 0.85

1.73 ± 0.54

EF

1.55 ± 1.75

1.54 ± 0.74

1.48 ± 0.32

REF

1.55 ± 1.75

7.69 ± 3.71

29.58 ± 6.38

CCR

0.50 ± 0.01

0.50 ± 0.00

0.50 ± 0.00

MCC

0.01 ± 0.02

0.01 ± 0.02

0.02 ± 0.02

CKC

0.01 ± 0.02

0.01 ± 0.01

0.01 ± 0.01

SEN

0.02 ± 0.02

0.02 ± 0.01

0.01 ± 0.00

SPE

0.99 ± 0.00

0.99 ± 0.00

0.99 ± 0.00

PRE

0.02 ± 0.02

0.08 ± 0.04

0.30 ± 0.06

ACC

0.98 ± 0.00

0.94 ± 0.00

0.80 ± 0.00

PM

0.58 ± 0.09

0.60 ± 0.04

0.62 ± 0.02

10

ROCE

1.50 ± 0.51

1.53 ± 0.24

1.62 ± 0.15

EF

1.49 ± 0.49

1.49 ± 0.22

1.44 ± 0.09

REF

14.88 ± 4.95

14.88 ± 2.16

28.73 ± 1.87

CCR

0.52 ± 0.03

0.53 ± 0.01

0.53 ± 0.01

MCC

0.02 ± 0.02

0.04 ± 0.02

0.07 ± 0.02

CKC

0.01 ± 0.01

0.03 ± 0.02

0.07 ± 0.01

SEN

0.15 ± 0.05

0.15 ± 0.02

0.14 ± 0.01

SPE

0.90 ± 0.00

0.90 ± 0.00

0.91 ± 0.00

PRE

0.01 ± 0.00

0.07 ± 0.01

0.29 ± 0.02

ACC

0.89 ± 0.00

0.86 ± 0.00

0.76 ± 0.00

PM

0.95 ± 0.02

0.98 ± 0.01

1.00 ± 0.00

1

20

ROCE

21.06 ± 7.29

46.82 ± 15.58

nana

EF

17.24 ± 4.92

13.94 ± 1.27

5.00 ± 0.00

REF

17.24 ± 4.92

69.71 ± 6.35

100.00 ± 0.00

CCR

0.58 ± 0.02

0.57 ± 0.01

0.53 ± 0.00

MCC

0.16 ± 0.05

0.30 ± 0.03

0.20 ± 0.00

CKC

0.16 ± 0.05

0.22 ± 0.02

0.08 ± 0.00

SEN

0.17 ± 0.05

0.14 ± 0.01

0.05 ± 0.00

SPE

0.99 ± 0.00

1.00 ± 0.00

1.00 ± 0.00

PRE

0.17 ± 0.05

0.70 ± 0.06

1.00 ± 0.00

ACC

0.98 ± 0.00

0.95 ± 0.00

0.81 ± 0.00

PM

0.90 ± 0.01

0.93 ± 0.00

1.00 ± 0.00

10

ROCE

9.30 ± 0.57

13.38 ± 0.59

1612.74 ± 529.71

EF

8.58 ± 0.49

8.26 ± 0.22

4.99 ± 0.01

REF

85.82 ± 4.86

82.60 ± 2.15

99.84 ± 0.18

CCR

0.88 ± 0.02

0.88 ± 0.01

0.75 ± 0.00

MCC

0.25 ± 0.02

0.56 ± 0.02

0.67 ± 0.00

CKC

0.14 ± 0.01

0.52 ± 0.02

0.61 ± 0.00

SEN

0.86 ± 0.05

0.83 ± 0.02

0.50 ± 0.00

SPE

0.91 ± 0.00

0.94 ± 0.00

1.00 ± 0.00

PRE

0.09 ± 0.00

0.41 ± 0.01

1.00 ± 0.00

ACC

0.91 ± 0.00

0.93 ± 0.00

0.90 ± 0.00

  1. In the case of bad model quality (λ = 1), the metrics most sensitive to variations in the R a value include the REF, PRE and ACC metrics, and also the CKC metric in the case of a large cutoff value of χ = 10%. This dependency is not so outspoken for the PM metric, except in the case when a very bad model is combined with a low cutoff value (χ = 1%). In cases with better model quality (λ = 20), significant dependencies are observed for the ROCE, EF, REF, MCC, CKC, SEN, PRE and ACC metrics, while the PM, CCR and SPE metrics are more stable. The metric that is least sensitive to variations in the R a value, irrespective of the underlying model quality or cutoff threshold, is the CCR metric
  2. aIn this case the ROCE metric could not be calculated from Eq. 10 since (N s  − n s ) is equal to 0