Table 4 Comparative results with SVM classifier and Random Forest
From: Large-scale prediction of protein ubiquitination sites using a multimodal deep architecture
Model | Input | Metrics | |||
|---|---|---|---|---|---|
Accuracy | Sensitivity | Specificity | MCC | ||
SVM | One hot vector | 59.65% | 46.69% | 61.42% | 0.054 |
Physico-chemical property | 57.36% | 43.84% | 59.22% | 0.051 | |
PSSM | 55.71% | 44.29% | 57.84% | 0.047 | |
Merged | 56.92% | 44.34% | 58.97% | 0.049 | |
Random Forest | One hot vector | 57.27% | 45.01% | 58.94% | 0.026 |
Physico-chemical property | 56.55% | 47.40% | 57.80% | 0.034 | |
PSSM | 54.19% | 44.98% | 56.32% | 0.021 | |
Merged | 56.52% | 46.36% | 58.83% | 0.024 | |
Our deep architecture | One hot vector | 64.15% | 64.41% | 64.08% | 0.189 |
Physico-chemical property | 61.84% | 60.97% | 61.95% | 0.151 | |
PSSM | 56.82% | 58.73% | 56.57% | 0.099 | |
Merged | 66.43% | 66.67% | 66.40% | 0.221 | |