Table 3. Protocol of "CNS/not-CNS" classification of compounds by intuitive approaches and statistical methods. TP-correct recognition of "CNS", FN-incorrect recognition of "CNS", TN-correct recognition of "not-CNS", EP-misdiagnosis, "not CNS", SE-sensitivity (TP / TP + EN), SP- (TN / TN ++ EP), ACC- accuracy (TP + TN) / (TP + EN + TN + EP).

Approach

Desciptors

Sets

TP

FN

TN

FP

SE

SP

ACC

“Rule of 5” 1997 [70]

MW  >  500, MlogP > 4.15, HBD > 5, HBA > 10

Training

External

397

47

103

3

198

11

302

39

0.794

0.940

0.396

0.22

0.595

0.580

Waterbeemd [81]

MW < 450, PSA < 90Å2, logD 1-4

Training

External

219

30

281

29

416

43

84

7

0.438

0.600

0.832

0.860

0.635

0.730

Norinder [71]

N+O ≤ 5

Training

External

496

50

4

0

59

5

441

45

0.992

1.000

0.118

0.100

0.555

0.550

Norinder [71]

ClogP-(N+O) > 0

Training

External

345

36

155

14

304

34

196

16

0.690

0.720

0.608

0.680

0.649

0.700

Raub [74]

clogP < 4, TPSA 40-80Å2

Training

External

178

23

322

27

394

37

106

13

0.356

0.460

0.644

0.740

0.572

0.600

Hitchcock [78]

PSA < 90Å2, HBD < 3, logP 2-5, logD2-5, W < 500

Training

External

69

15

431

35

432

46

48

4

0.138

0.300

0.864

0.820

0.501

0.610

Wager [75]

MPO score ≥ 4.0 / < 4.0

Training

External

342

34

158

16

256

18

244

32

0.684

0.680

0.512

0.360

0.598

0.520

LR

maxQ-,maxCa
maxCa*maxCd

Training

CV

External

370

369

42

130

131

8

339

333

39

161

167

11

0.740

0.738

0.840

0.678

0.666

0.780

0.709

0.702

0.810

LR

maxQ-,maxC maxCa*maxCd
MW, HBD,NCC,logD

Training

CV

External

382

384

43

118

116

7

340

336

38

160

164

12

0.764

0.768

0.860

0.680

0.672

0.760

0.722

0.720

0.810

RF

*

Training

CV (out-of-bag)

External

500

412

47

0

88

3

500

378

35

0

122

15

1.000

0.824

0.940

1.000

0.756

0.700

1.000

0.790

0.820

8SVM

**

Training

External

448

46

52

4

403

38

97

12

0.896

0.920

0.806

0.760

0.851

0.840

Note. *α maxQ+ maxQ- ∑Q+ ∑Q+/α maxEa maxCa maxEamaxEd ∑Ca/α ∑Cd/α NRB NCC logDD pKa PSAed TPSA(N,O) MlogP;
**maxQ+ maxQ% ∑Q+ ∑Q+ /α maxEa maxCa maxCd maxEa*maxEd maxCa*maxCd ∑Ed ∑Ead ∑Ea /α ∑Ed/α ∑Ca /α ∑Cad/α MW HBD NRB NCC logPP logDD pKa PSAed TPSA(NO) MlogP AlogP