journal menu
![[Figure 3]](dez5002fig3mag.jpg)
|
|
Figure 3
PETIMOT's performance on the test set and comparison with other methods. PETIMOT-default is compared with AlphaFlow, BioEmu and the NMA on test824. For each protein, we evaluate the four motions predicted by PETIMOT or inferred from AlphaFlow/BioEmu-predicted ensembles against the four ground-truth motions. We allow the NMA more flexibility by considering the ten lowest frequency modes. (a) Minimum LS error, computed for the best-matching pair of predicted and ground-truth motions. (b) Minimum magnitude error. (c) Global SS error, reflecting ground-truth subspace coverage. (d) Distributions of maximum minimum r.m.s.d.: within each generated ensemble, we identify the conformations closest to the reference structures and compute the maximum r.m.s.d. among them. If this value is below 2.5 Å then we consider that all reference structures are covered by the ensemble. Single Struct. refers to the input structure given to PETIMOT. Orig. stands for the original ensembles generated directly from BioEmu and AlphaFlow, while Iso. stands for isotropic ensembles generated from motions predicted by PETIMOT or inferred from BioEmu/AlphaFlow original ensembles. |
![[Figure 3]](dez5002fig3.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
access
