A Systematic Evaluation of Feature Selection and Classification Algorithms Using Simulated and Real miRNA Sequencing Data
Table 3
Sensitivity and specificity of the seven algorithms in different settingsa.
Scenario
baySeq
DESeq
edgeR
Lasso
Rank sum
PSODT
RF
Bonb
Bonb
Bonb
Bonb
s2n
A1
0.68/1.00
0.16/1.00
0.97/0.96
0.43/1.00
0.99/0.93
0.68/1.00
0.62/0.99
0.92/0.95
0.16/1.00
0.10/0.99
0.60/1.00
A2
0.74/1.00
0.18/1.00
0.97/0.96
0.37/1.00
0.98/0.92
0.63/1.00
0.49/0.99
0.92/0.95
0.16/1.00
0.11/0.95
0.70/0.98
A3
0.77/1.00
0.19/1.00
0.95/0.94
0.32/1.00
0.98/0.91
0.57/1.00
0.32/1.00
0.92/0.95
0.16/1.00
0.14/0.90
0.71/0.97
A4
0.77/1.00
0.17/1.00
0.93/0.93
0.27/1.00
0.96/0.89
0.50/1.00
0.23/1.00
0.92/0.95
0.16/1.00
0.18/0.86
0.70/0.95
A5
0.76/1.00
0.16/1.00
0.90/0.90
0.22/1.00
0.95/0.86
0.43/1.00
0.18/1.00
0.18/1.00
0.92/0.95
0.17/1.00
0.70/0.93
Mean of significant variables
B1
0.05/1.00
0.00/1.00
0.41/0.95
0.01/1.00
0.52/0.92
0.04/1.00
0.13/0.99
0.40/0.95
0.01/1.00
0.12/0.90
0.33/0.93
B2
0.45/1.00
0.03/1.00
0.81/0.95
0.12/1.00
0.88/0.92
0.27/1.00
0.30/1.00
0.77/0.95
0.05/1.00
0.13/0.90
0.57/0.95
B4
0.91/1.00
0.41/1.00
0.99/0.94
0.53/1.00
0.99/0.91
0.78/1.00
0.32/1.00
0.97/0.95
0.31/1.00
0.14/0.91
0.79/0.98
B5
0.96/1.00
0.62/1.00
1.00/0.94
0.66/1.00
1.00/0.90
0.90/1.00
0.32/1.00
0.99/0.95
0.45/1.00
0.14/0.91
0.83/0.98
Dispersion parameter of significant variables
C1
1.00/1.00
0.87/1.00
1.00/0.92
0.57/1.00
1.00/0.89
0.92/1.00
0.37/1.00
1.00/0.95
0.97/1.00
0.26/0.95
0.97/1.00
C2
0.89/0.99
0.38/0.94
0.98/1.00
0.40/0.94
0.99/1.00
0.75/0.97
0.35/0.93
1.00/1.00
0.61/0.96
0.14/0.90
0.86/0.90
C4
0.71/1.00
0.14/1.00
0.90/0.95
0.29/1.00
0.92/0.92
0.36/1.00
0.24/0.99
0.46/0.95
0.01/1.00
0.14/0.90
0.52/0.95
C5
0.73/1.00
0.28/1.00
0.73/0.96
0.29/1.00
0.71/0.93
0.16/1.00
0.00/1.00
0.23/0.95
0.00/1.00
0.13/0.90
0.44/0.94
The conditions where the mean = 20, dispersion parameter = 1, and s2n = 0.1 are the same. Each cell includes the sensitivity and specificity. Bon indicates a result using the Bonferroni correction.
