*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
Jimson Wong wrote:
Through labourous manual rebuilding and with the help of DM solvent flipping + NCS averaging map the Rfree and Rfactor are both 0.271 at the moment, with only 200 water molecules along with 680 residues.
The most fundamental question I can see here is why are your R-factor and R-free the same ???? I'm assuming that you're at 2.5 Angstrom or better, given all those solvent molecules. But since you've added solvent molecules I assume that you've refined the structure somewhat already. That would mean that your R and R-free should NOT be the same unless you somehow rebuilt the entire structure (dedicated but not necessary) or started adding water molecules before you've actually refined the structure at all (unwise unless you've got nearly perfect experimental phases and better than 1.6 Angstroms resolution, in which case ARP/wARP could have built the whole thing for you while you drank coffee).
You'd expect an R-free - R of at least 4%, unless you have a lot of NCS and unless you are at very high resolution.
I assume you are using NCS restraints during refinement ? Although this won't change things all that much.
I could add more water molecules to drive down the Rfree,
A better philosophy would be to add more water molecules to improve the interpretation of the electron density map, but that's dogma for another email.
Would any experienced CNS users know what it means when the Rfree starts to go up when one uses minimize.inp?
My best guess is that your R and R-free are not independent of each other yet: you've changed the R-free set during refinement, or you've not refined your structure yet.
If my best guess is wrong: the more general case is that you've made a bad choice for WA (the weight on the amplitudes relative to the geometry). Try setting it to -1 and see what CNS chooses. Try half that value. Try running the script optimize_wa.inp and see what gives the lowest R-free with reasonable geometry. Look at the structure before and after refinement to see what has changed (e.g. regions that had bad contacts that got mangled when CNS relieved the atom contact collisions ahead of everything else).
But more details would help improve the guess. Phil Jeffrey Crystallography Facility Manager Princeton University
