Table 1.
Summary of prediction accuracy for the proposed method and other methods
| (A) Internal motion | ||||||
|---|---|---|---|---|---|---|
| Method | MAE | CC | AUC | |||
| avg. | avg. | SD | 1 | 0 | -1 | |
| proposed method (PHD & RVPnet) | 0.621 | 0.482 | 0.183 | 0.743 | 0.765 | 0.768 |
| proposed method (psipred & sable) | 0.605 | 0.525 | 0.197 | 0.759 | 0.786 | 0.791 |
| naïve model (PHD) | 0.988 | 0.248 | 0.161 | 0.633 | 0.653 | 0.688 |
| naïve model (psipred) | 0.952 | 0.293 | 0.175 | 0.666 | 0.672 | 0.708 |
| PROFbval | 0.743 | 0.367 | 0.199 | 0.711 | 0.693 | 0.698 |
| POODLE-S | - | - | - | 0.713 | 0.730 | 0.755 |
| FlexPred | - | - | - | 0.751 | 0.741 | 0.768 |
| (B) External motion | ||||||
| Method | MAE | CC | AUC | |||
| avg. | avg. | SD | 1 | 0 | -1 | |
| proposed method (PHD & RVPnet) | 0.571 | 0.541 | 0.188 | 0.770 | 0.777 | 0.81 |
| proposed method (psipred & sable) | 0.542 | 0.597 | 0.209 | 0.806 | 0.806 | 0.843 |
| naïve model (PHD) | 0.970 | 0.262 | 0.135 | 0.650 | 0.661 | 0.697 |
| naïve model (psipred) | 0.929 | 0.320 | 0.145 | 0.685 | 0.681 | 0.733 |
| PROFbval | 0.608 | 0.547 | 0.167 | 0.785 | 0.784 | 0.844 |
| POODLE-S | - | - | - | 0.756 | 0.783 | 0.841 |
| FlexPred | - | - | - | 0.791 | 0.777 | 0.817 |
Herein, avg. and SD respectively signify the average and standard deviation. The highest scores in each criterion are underlined. Here, -1, 0, and 1 are threshold values used to discriminate rigid and flexible classes for plotting the ROC curve. PROFbval, POODLE-S, and FlexPred were performed using the default parameters. Here, POODLE-S and FlexPred respectively produced disorder probability and probability of flexible label. Therefore, MAE and CC were not calculated. The AUC for POODLE-S and FlexPred were calculated using data of normalized disorder probabilities and probabilities of flexible labels.