> Could anyone offer me an explanation for the following, or suggest
> Recently, I solved a structure of a coiled coil, chemeric protein
> construct that contains the GCN4 leucine zipper. I built models of
> contruct, but could not find a molecular replacement solution with
> The next thing I did was this: I performed molecular replacement with
> the GCN4 leucine zipper alone with AMoRe and found a solution with
> rotation but wrong translation. I then extended and rebuilt the model
> and refined it with CNS to convergence.
> In the next step, I performed molecular replacement again, this time
> with the
> "refined" structure, and found the correct solution with two molecules
> in the asymmetric unit. Now here is the question, how could the
> refinement of a model with correct orientation and incorrect position
> improve the model so that I could find the correct solution with the
> "refined" model later? Is anyone aware of any reference that
> mentioned this
> kind of method before?
I've seen bad solutions with the right rotation used to refine a
model successfully, even with the wrong spacegroup. It probably depends
a lot on the spacegroup, and how the translation is wrong, like placing
a molecular 2-fold onto a 2(1) axis.
In what way is the translation wrong? Are you sure it isn't correct,
but for a different origin choice?