Prediction of Enzyme Mutant Activity Using Computational Mutagenesis and Incremental Transduction
Table 1
One-shot learning.
Dataset
Algorithm
Avg. Acc.
St. Dev.
Sensitivity
Specificity
HIV-1
AdaBoost
68.84
4.77
0.87
0.63
LogitBoost
75.93
3.90
0.72
0.86
SVM
68.65
3.50
0.66
0.75
RF
79.28
1.96
0.88
0.76
DT
77.57
1.21
0.73
0.81
T1 NN
73.13
2.57
0.69
0.76
T1 RF
74.64
3.24
0.65
0.74
T4
AdaBoost
85.10
0.14
0.98
0.11
LogitBoost
85.65
0.31
0.97
0.20
SVM
86.88
0.24
0.99
0.17
RF
87.12
0.44
0.97
0.30
DT
85.33
0.56
0.93
0.34
T1 NN
75.46
6.99
0.80
0.46
T1 RF
85.02
7.44
0.94
0.35
LAC
AdaBoost
60.53
0.31
0.99
0.12
LogitBoost
71.88
0.58
0.91
0.48
SVM
72.15
0.16
0.88
0.52
RF
80.80
0.37
0.86
0.75
DT
78.71
0.34
0.83
0.74
T1 NN
65.23
3.58
0.76
0.39
T1 RF
77.73
3.64
0.78
0.77
Results of the AdaBoost, LogitBoost, SVM, random forest, decision tree, and transduction T1 algorithms. Using one-shot learning, no selectivity, and 4-fold cross-validation for the HIV-1 dataset and 10-fold cross-validation for T4 and LAC datasets.