journal menu
![[Figure 2]](rr5237fig2mag.jpg)
|
|
Figure 2
Dependence of mr_rosetta performance on pLDDT cutoff for AlphaFold models of Ceα2/β5. Molecular replacements were performed with AlphaFold models of Ceα2/β5 produced in monomer mode with varied pLDDT cutoff scores, which were then elaborated by the Rosetta–AutoBuild procedure of mr_rosetta. For each pLDDT cutoff, histograms are shown for the residue count in the Ceα2/β5 search model (gray), the mr_rosetta constructed model of two copies of Ceα2/β5 in the asymmetric unit (orange) and the mr_rosetta constructed model having Cα deviations less than 2 Å from the fully refined model (blue). The yellow line is at the residue count for one copy of the fully refined Ceα2/β5 model (187). It was surprising that the autobuilt models from pLDDT cutoffs at 60, 65 and 70 showed were so dissimilar to the refined structure (no blue bars). These all had the same Phaser solutions; however, in these cases the model extensions from mr_rosetta built into crystal symmetry mates, fitting into correct backbone geometry but completely out of register, introducing deletions and gaps. As a result, none of the autobuilt structures met the superposition criteria. |
![[Figure 2]](rr5237fig2.jpg)
 | STRUCTURAL BIOLOGY |
ISSN: 2059-7983
