|
|
|
Figure 2
PETIMOT prediction visualization, evaluation and comparison with other methods. (a, b) Prediction for B. subtilis xylanase A (PDB entry 3exu, chain A). (a) The predicted and ground-truth vectors are in blue and red, respectively (LS = 0.15). (b) Trajectory generated by deforming the protein structure along the predicted motion. (c, d) Success rates computed as the proportions of test proteins with mimimum LS below 0.2 (dark), 0.4 (mild) and 0.6 (light). (c) Comparison of PETIMOT's default (blue), stringent (tomato red) and 5folds (gold) models. Test sets avoid data leakage at different levels. Test824: less than 80% sequence similarity to any collection used for training. Full dataset: folds are defined based on strict sequence and structural similarity filters. Test734: less than 30% sequence similarity to training collections. Test474: less than 30% sequence similarity and no significant structural similarity to training collections. (d) Comparison of PETIMOT's default model (black) with AlphaFlow (blue), BioEmu (magenta) and the NMA (tomato red) on test824. |

journal menu![[Figure 2]](dez5002fig2.jpg)
access


