RAVINDRA D MAKDE wrote:
Hi everyone,
I would like to have commnets from experts on the following issues.
I am refining DNA-protein complex at 3 A resolution and my current Rfree is 21.5 % (phenix.refine).
1. I am getting difference map peaks (in Coot) of magnitude less than -5 sigmas (0.24 e/A3) (~10 or more peaks) and equal number of positive peaks of same magnitudes. The positive peaks hardly have 2fo-fc of more than 1.0 sigma level. I presumed this well could be due to bulk solvent or improper mask. Therefore I optimized the mask parameters (by giving option under "General refinement parameters" in phenix.refine GUI). I could get my R free lower as expected but I still got these peaks back. Although I am not absolutely sure, but positive peaks are more in polar pocket of protein and negative peaks are more in non-polar and aromatic pockets. I have PEG, Na acetate in my crystal soup. What these negative peaks represent for? I suspect that negative peaks are from the bulk-solvent mask, as you suspected. Some proteins have hydrophobic cavities that are big enough to make a small void that ends up with a bit of bulk solvent density. Optimizing bulk solvent parameters does not necessarily help, because similar sized cavities may in fact have a bit of actual solvent density.
It may be possible to put zero-occupancy dummy atoms on those sites to screen out the bulk solvent mask, but I don't know how the PHENIX bulk mask handles partial occupancy.
2. During addition of atoms like Na+, Cl- in map, do one need to careful about of the its coordination valency in surrounding pocket. ?
Definitely. The PDB is full of misidentified ions, even in active sites at resolutions much better than 3.0. Many Polymerase Beta structures have an Mg++ ion in the model, where the true identity is Na+. It is easy for ions to have partial occupancy with waters or other ions, or some disorder due to the water ligands. Of course, there are probably even more invalid waters in the PDB than anything else. My opinion is that it is better to make your best guess rather than just leave in phony waters, and put remarks to the deposited PDB file for anything that is uncertain. Unfortunately, most PDB consumers don't read remarks, so we would benefit from a standard way to flag qualitative, author-defined uncertainties. Joe Krahn