Skip to main content
. 2018 Oct 26;9:2551. doi: 10.3389/fmicb.2018.02551

Table 7.

The performance of SVM based models developed using different features on additional dataset.

Features (parameters) Mod_AMP_similar Dataset
Sen Spc Acc MCC AUROC
Atom composition (g = 0.1, c = 9, j = 1) 89.47 43.68 66.58 0.37 0.77
Diatom composition (g = 0.05, c = 15, j = 2) 88.42 71.58 80.00 0.61 0.88
2D descriptors (g = 0.1, c = 1, j = 1) 84.74 32.63 58.68 0.20 0.66
Fingerprints (g = 0.001, c = 4, j = 1) 93.16 87.37 90.26 0.81 0.97
Hybrid features (2D + fingerprints) (g = 0.005, c = 7, j = 2) 84.74 58.95 71.84 0.45 0.81
N100C100 Binary profile (only atoms) (g = 0.01, c = 6, j = 1) 90.51 89.44 89.66 0.80 0.97
N100C100 Binary profile (only symbols) (g = 0.005, c = 7, j = 2) 76.98 91.10 84.21 0.60 0.94
N200C200 Binary profile (atom + symbols) (g = 0.005, c = 1, j = 2) 89.29 89.12 89.20 0.78 0.96

Sen, Sensitivity; Spc, Specificity; Acc, Accuracy; MCC, Matthew’s Correlation Coefficient; AUROC, Area Under the Receiver Operating Characteristic curve; N100C100, first 100 elements form N-terminus and C-terminus, respectively; N200C200, first 200 elements form N-terminus and C-terminus, respectively.