Table 2. The parameters of training and testing linear regression models on sets for individual receptors and their combinations.

 

Model based on set:

Parameter

M1

M2

M3

M4

M5

M1, M4, M5

M1, M3, M4, M5

M1, M2, M4, M5

M1, M2, M3, M4, M5

(неч)

M1, M2, M3, M4, M5

(чет)

M1, M2, M3, M4, M5

 

The total number of observations in the original setбщее число наблюдений в исходной выборке

34

86

108

28

28

90

198

176

-

-

284

The number of observations selected by the position obtained after docking

24

70

62

23

20

67

129

137

100

99

199

The range of pKi values after selection by docking

7.3-9.92

5.3-9.9

5.6-9.9

6.2-9.8

6.8-10

6.2-10

5.6-10

5.3-10

5.3-10

5.3-9.9

5.3-10

The number of variables in equation

9

9

9

9

9

11

12

12

13

13

13

R2 of leaning

0.49

0.71

0.66

0.92

0.61

0.7

0.66

0.78

0.74

0.70

0.71

Q2 in leave-one-out procedure

0.18

0.58

0.53

0.88

0.1

0.4

0.58

0.72

0.65

0.67

0.61

MSE of leaning

0.37

0.55

0.4

0.07

0.22

0.2

0.35

0.35

0.37

0.44

0.42

The averaged R2 R2 of learning in random mixing (10 attempts)

0.45

0.09

0.12

0.3

0.51

0.16

0.08

0.08

0.12

0.14

0.07

R2 of prediction for set M1

-

0.03

0.08

0.45

0.53

-

-

-

-

-

-

R2 of prediction for set M2

0.19

-

0.38

0.7

0.19

0.25

0.5

-

-

-

-

R2 of prediction for set M3

0.32

0.48

-

0.64

0.45

0.44

 

0.45

 

 

 

R2 of prediction for set M4

0.31

0.52

0.46

-

0.43

-

-

-

-

-

-

R2 of prediction for set M5

0.41

0.06

0.01

0.49

-

-

-

 

 

 

 

R2 of prediction for half of common set (M1, M2, M3, M4, M5, odd)

-

-

-

-

-

-

-

-

-

0.7

-

R2 of prediction for half of common set (M1, M2, M3, M4, M5, even)

-

-

-

-

-

-

-

-

0.68

-

-

Note. "-" - not calculated or not applicable. The set labeled M1 (or M2 or M3 or M4 or M5) is a set for individual receptor. When set name listed separated by commas (M1, M4, M5), these is combined set that include data for the corresponding receptors.