Hi Lithlin, I guess I have an idea about what's happening. When you add dummy atoms (unstructured or as some model) with zero occupancy and ask the mask to not set up bulk-solvent in that region, the (difference) maps you get are contrasted in the region of dummy atoms; that is density becomes systematically different in that region. Whether that is because of genuine feature or classical model bias isn't clear. Perhaps that's not limited to difference maps.. Pavel On 4/21/13 7:50 AM, Jinzhong Lin wrote:

On 04/21/2013 01:22 AM, Nathaniel Echols wrote:

...

On Sat, Apr 20, 2013 at 7:40 PM, Jinzhong Lin wrote:

...

I suspect that model will still be used to calculate the Fc even though the occupancy is zero, so the map will be severely model biased. Technically, they will be included in the Fc calculation in our implementation, but since the occupancy is zero, their contribution to Fc will be as well. So no, model bias should not be a problem.

-Nat _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb

It is an impression I got from two of my colleagues both of who had spent some time playing with bulk solvent model.

One was trying to add a tRNA into his structure. He could see some densities for the tRNA, so he put a tRNA model there, set the occupancy to zero and adjusted the setting. Since the resolution is only 7 A, he only tried rigid body refinement. Surprisingly, the resulting map showed an unreasonably good density for the tRNA. He put the tRNA model in a completely different position, he also got the density back in that wrong position.

The other person was doing the same thing except this time his model he wanted to put in is a small molecule. The resolution is 4A. He found him in the same situation.

I am not sure if they have done things correctly. I am going to do this on my own, I will keep you all updated. _______________________________________________ phenixbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/phenixbb