In this publication here, the comparative model algorithm of rosetta is explained:
However, in case of having a density to guide the model completion, how does the density factor in this algorithm?
See the tutorials at https://www.rosettacommons.org/demos/latest/public/electron_density_structure_refinement/structure_refinement for details on how do do this.
The way density is used in these case is that we add an additional scoreterm to the energy function, measuring how good a fit the structure has to the density. That is, instead of just optimizing over the "goodness of structure" (the regular Rosetta energy function), we optimize over both "goodness of structure" and "goodness of fit". The hope is that there's one clear energy well (the native structure) which is simultaneously fits into the density well and has a reasonable protein structure.
Thanks a lot. How do you measure the goodness of fit, fsc score? How do you calculate it?
Goodness of fit is measured using real-space correlation.
To expand on rmoretti's point, density-fit is used as an additional term in all 3 stages of the algorithm:
Thanks a lot. Is real space correlation also used for fragment searching when doing de novo? Do you only use an integrated FSC from 10 to the resolution of the data as previously explained in the refinedment tutorial to assess the accuracy of a refined structure?