Evaluation and Identification of the Neuroprotective Compounds of Xiaoxuming Decoction by Machine Learning: A Novel Mode to Explore the Combination Rules in Traditional Chinese Medicine Prescription
Table 3
Performance of single classification models for the training set (5-fold cross-validation result) and the test set (validation result using external test set) using different combinations of molecular properties.
No.
Model
Descriptors
5-fold cross-validation result
Validation result using external test set
SE
SP
PPV
MCC
SE
SP
PPV
MCC
1
RF-a1
32
0.682
0.970
0.912
0.750
0.675
0.975
0.915
0.718
2
RF-b1
53
0.621
0.973
0.903
0.750
0.650
0.981
0.915
0.716
3
RF-c1
79
0.667
0.992
0.927
0.823
0.690
0.980
0.922
0.742
4
K-NN-a1
32
0.682
0.905
0.861
0.710
0.695
0.926
0.880
0.622
5
K-NN-b1
53
0.621
0.936
0.873
0.710
0.599
0.931
0.865
0.557
6
K-NN-c1
79
0.636
0.936
0.876
0.760
0.609
0.931
0.867
0.565
7
Tree-a1
32
0.621
0.943
0.879
0.680
0.635
0.942
0.881
0.609
8
Tree-b1
53
0.667
0.939
0.885
0.566
0.624
0.914
0.856
0.545
9
Tree-c1
79
0.682
0.947
0.894
0.780
0.711
0.943
0.897
0.670
10
AB-a1
32
0.682
0.920
0.873
0.659
0.731
0.900
0.867
0.603
11
AB-b1
53
0.667
0.936
0.882
0.741
0.695
0.902
0.860
0.578
12
AB-c1
79
0.621
0.928
0.867
0.824
0.741
0.941
0.901
0.687
13
NB-a1
32
0.766
0.795
0.790
0.483
0.758
0.746
0.748
0.421
14
NB-b1
53
0.777
0.904
0.879
0.644
0.682
0.894
0.852
0.555
15
NB-c1
79
0.761
0.908
0.879
0.640
0.712
0.905
0.867
0.598
16
RF-a2
26
0.805
0.994
0.956
0.860
0.655
0.991
0.924
0.710
17
RF-b2
50
0.839
0.994
0.963
0.882
0.690
0.983
0.924
0.675
18
RF-c2
65
0.747
0.997
0.947
0.830
0.724
1.000
0.945
0.761
19
K-NN-a2
26
0.793
0.960
0.926
0.766
0.724
0.957
0.910
0.574
20
K-NN-b2
50
0.747
0.945
0.906
0.702
0.724
0.957
0.910
0.585
21
K-NN-c2
65
0.782
0.951
0.917
0.739
0.793
0.957
0.924
0.597
22
Tree-a2
26
0.759
0.971
0.929
0.769
0.655
0.966
0.903
0.601
23
Tree-b2
50
0.805
0.937
0.910
0.726
0.448
0.983
0.876
0.629
24
Tree-c2
65
0.782
0.963
0.926
0.765
0.793
0.966
0.931
0.656
25
AB-a2
26
0.805
0.937
0.910
0.726
0.655
0.957
0.897
0.602
26
AB-b2
50
0.828
0.931
0.910
0.732
0.793
0.948
0.917
0.621
27
AB-c2
65
0.851
0.951
0.931
0.788
0.828
0.974
0.945
0.570
28
NB-a2
26
0.839
0.951
0.929
0.780
0.724
0.871
0.841
0.551
29
NB-b2
50
0.816
0.966
0.936
0.796
0.621
0.914
0.855
0.542
30
NB-c2
65
0.885
0.943
0.931
0.795
0.712
0.905
0.867
0.598
1-15: neuroprotective models against hypoxia-induced neurotoxicity (NIN models). 16-30: neuroprotective models against H2O2-induced neurotoxicity (NHN models). a: models built by DS_2D descriptors. b: models built by MOE_2D descriptors. c: models built by DS_MOE 2D descriptors.