Figure 3.

An overview of trRosetta-predicted domains. (a) The top three models from trRosetta for ten representative domains indicate a tight convergence of modeling. The identity of the domains follows the coloring in Fig. 2 ▸(a). Domains from FANCB, FANCE and FANCC are shown in the top row, while those from FAAP100 and FANCG are shown in the bottom row. (b) Several examples of trRosetta models docked into density before refinement, showing the role that the map plays in the validation and selection of models. From left to right [the colors match those in Fig. 2 ▸(a)]: the helical repeats of FANCC, the N-terminal repeats of FANCE, the α/β domain of FANCB and the β-sandwich of FAAP100. (c, d) Two examples illustrating the importance of domain segmentation when docking trRosetta-generated models. (c) The trRosetta model of FANCG (magenta) poorly matches the final structure (green); segmenting this model into two domains (red and blue) shows a much better match, as the individual domain structures are accurate, even though their relative orientation is not. (d) Similarly, a trRosetta prediction of the FANCB β-sandwich–α/β domain (pink) is dissimilar from the final structure (blue); splitting it into domains (brown and green) shows good overall agreement. (e) trRosetta models (blue) generally fit the map well, although some refinement was necessary to maximize agreement with the density (orange).