Table 4. Comparison of the prediction performance.
| Our Results | Saha S, et al. [7] | Song K, et al. [28] | Ning LW, et al. [31]5 | Deng J, et al. [12] | Gustafson AM, et al. [9] | Gerdes SY, et al. [32] | Joyce AR, et al. [33] | Plaimas, et al.[11] | Ye YN, et al. [34] | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GN | GP | FULL (GN+GP) | S. cerevisiae | E. coli2 | B. subtilis3 | Max | E. coli | M. pulmonis | Combine | E. coli | E. coli | E. coli | E. coli | E. coli | E. coli | |||
| SVM | KNN1 | (Min)4 | BLAST6 | CEG_MATCH6 | ||||||||||||||
| Sensitivity | 0.709 | 0.733 | 0.715 | 0.768 | 0.742 | 0.760 | 0.792 | 0.904 | / | / | / | 0.73 | 0.52 | 0.68 | 0.26 | / | / | / |
| (0.609) | ||||||||||||||||||
| Specificity | 0.733 | 0.786 | 0.736 | / | / | 0.867 | 0.858 | 0.926 | / | / | / | 0.92 | 0.96. | 0.88 | 0.25 | / | 0.431 | 0.60 |
| (0.778) | (0.345) | (0.694) | ||||||||||||||||
| AVE | 0.721 | 0.760 | 0.726 | / | / | 0.814 | 0.825 | 0.898 | / | / | / | / | / | / | / | / | / | / |
| (0.735) | ||||||||||||||||||
| AUC | 0.789 | 0.838 | 0.794 | 0.81 | 0.81 | 0.866 | 0.870 | 0.937 | 0.82 | 0.74 | 0.76 | / | / | / | / | 0.81 | / | / |
| (0.804) | (0.75) | |||||||||||||||||
| ACC | 0.731 | 0.763 | 0.734 | 0.741 | 0.734 | 0.904 | 0.903 | 0.960 | / | / | / | / | / | / | / | / | 0.694 | 0.712 |
| (0.813) | (0.677) | (0.701) | ||||||||||||||||
| PPV | 0.226 | 0.330 | 0.243 | 0.731 | 0.730 | 0.709 | 0.673 | 0.942 | / | / | / | 0.44 | 0.53 | 0.33 | 0.42 | / | / | / |
| (0.435) | ||||||||||||||||||
| Number of feature | 37 | 38 | 40 | 13 | 13 | 4947 | 4947 | / | 158 | 158 | 158 | 13 | 28 | / | / | 338 | / | / |
| Number of object | 21 | 10 | 31 | 1 | 1 | 11 | 11 | / | 1 | 1 | 16 | 1 | 1 | 1 | 1 | 3 | 169 | 169 |
1 k-nearest neighbor (KNN) method
2 Date of E. coli was used for training. The data of the other 11 objects were used as test set, and the results were averaged.
3 Date of B. subtilis was used for training. The data of the other 11 objects were used as test set, and the results were averaged.
4 The maximum and the minimum values of the prediction results.
5 Results based on cross validation were chosen for comparison.
6 BLAST: Identity >50. CEG_MATCH: K = 3.
7 The features were the 93’ Z-curve features (252 variables), orthologs values (187), and other DNA or amino acid sequence based features (55).
8 The features were the topology features (25) and the genomic and transcriptomic features (8).
9 Results of 16 objects were listed in [34].