journal menu
![[Figure 3]](lie5001fig3mag.jpg)
|
|
Figure 3
Averaged metrics depending on the sequence length for the different approaches (AlphaFold3, ab initio, deep-learning, template-based and challenge-best). Each point represents the metric averaged over the best models of each approach for a window of 25 nt from 25 to 750 nt. Ab initio methods group RNAJP, IsRNA1 and SimRNA, while template-based methods group Vfold-Pipeline, 3dRNA, RNAComposer, Vfold3D and MC-Sym. Deep-learning methods group trRosettaRNA and RhoFold. Metrics are computed for the RNA-Puzzles, CASP-RNA and RNASolo data sets. Challenge-best corresponds to the best results from either the RNA-Puzzles or CASP-RNA competitions but does not appear for the RNASolo data set. The metrics are RMSD, MCQ, TM-score, lDDT and INF-ALL. RMSD and MCQ are reversed to have the best values near the top and the worst values at the bottom. |
![[Figure 3]](lie5001fig3.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
access
