 |
 |
Kay Diederichs wrote: > Hi Joane, > > since you're using NCS I'd like to add that this is indeed a good > strategy, but you should make sure that you should probably not restrain > the B-factors of the well-defined domain and the not-so-well-defined > domains to be equal - they do differ and the refinement should probably > not change that! > So I'd say: adjust the weights such that the coordinates are > NCS-restrained, but not the B-factors (or only a little). Use R-free to > find the right values for the restraints. > > HTH, > > Kay I remember the late Jeroen Brandsen suggesting that B factors in different chains should not be restrained to be equal, but related in the same way. We don't restrain the coordinates to be equal, but rather related by a linear transform. Likewise B factors in chain B could be related to those in chain A by an equation like ADP(Bi) = m*ADP(Ai) + b. This would allow for one chain to be everywhere less ordered than the other, but in each chain the same atoms would have the highest B-factors. At the expense of only two linear parameters per additional chain. eab