This server uses normal modes to perform medium- or low-resolution structural refinement
of an initial model, using cryo-EM data as constraints; namely it will force the model to fit
the experimental map, which may have been obtained under other experimental conditions (i.e. bound to a ligand etc...).
This kind of refinement will enforce collective and large-amplitude movements in the initial model that are beyond
the reach of currently existing methods.
The refinement is carried out in reciprocal space with
respect to the normal mode amplitudes, using standard conjugate-gradient optimization.
If you have real-space data (i.e. a molecular envelope), you must first transform it
to structure factors in reciprocal space (Get Struct. Fact.).
See the Delarue and Dumas (2004) article for details.
Although in the original article we used points on a cubic lattice filling the purple envelope (closed form) to
calculate normal modes and deform them into the cyan envelope, it is of course possible to use the CAs atoms
of the molecule (citrate synthase in this case, a dimer of about 850 residues).
Click here to see the Movie for adenylate kinase.
We can easily refine the open form coordinates into
the closed form electron density at low resolution (15 or even 25 Angstrom). In that case, we get amplitudes for the first ten
lowest frequency normal modes very close to the ones obtained when one just minimizes the rmsd between the two forms
(see Figure on the right). The radius of Convergence of this refinement is therefore very large.