is the new data cryo and the old model derived from troom temp data ?
in that case, i nowadays use overall B-shift refinement after
the rigid-body refinement (refine bgroup, with b=(not hydrogen)
as the only group) and before any powell or sa
the reason is that the average b of the cryo model will be much
lower than that of the rt model which screws up positional refinement
(it will try to compensate for this but ince you don't refine bs
it iwill do it in the wrong way and rfree will be stuck; also
individual b-refinement may not converge properly if all that's needed
is a large shift of the average value); grouped b ref converges
very rapidly and has the advantages that (a) you're refining only
ONE extra parameter, and (b) you're not refining an overall B but
an overall B-factor *shift* (this means that regions with low bs
will remain refgions with low bs !)
in a few cases this cracked the problem
however, not in all ! i've heard of a case where rfree refused to
go donw until a bucket full of waters had been included in the model
perhaps we have to rethink the way we model proteins at cryo temps ?
--gerard
