Hi, I would suggest something more in the middle. Artifacts from the bulk solvent model tend not to be lumpy since it doesn't really have any Fourier coefficient of higher resolution than about 5 A. It sounds to me that you have something at partial occupancy. Contouring a map in units of "sigma" (which is really rmsd, but that is the topic for a flame war) can be very deceptive. As Herman said, there will always be peaks at 3 rmsd since it is a relative measure. One way to calibrate your contour level is to leave out something that you strongly believe is at full occupancy and compare the strength of your mystery density to this standard density. If the blob is only half as dense as the density of the known, good, object you know whatever it is it is there only about half the time. This means you not only have to build a model, but also refine its occupancy. Remember, if you see something binding at half occupancy, there is something else there the other half of the time. That might be bulk solvent, but more likely ordered water molecules will be binding in this pocket. This fact makes interpreting the density harder because you are looking at density that is the sum of a ligand and scattered water molecules, AND these water molecules likely make the same hydrogen bonds as your ligand. If you have a glycerol at half occupancy you will also have water molecules that bind in pretty much the same locations as the hydroxyl groups. Those atoms will appear at near full occupancy because of this superposition but the carbon atoms of the glycerol will appear at half occupancy. This can be a real mess, especially if the glycerol, itself, has multiple binding modes. How far down the rabbit hole of partial occupancy you want to go is a judgement call. Dale Tronrud On 8/24/2016 12:15 AM, [log in to unmask] wrote: > Dear Andrew, > > > > I suspect it is an artefact of the bulk solvent correction (or lack > thereof in the inaccessible pocket). What is the absolute value of the > difference density? Difference density gets by default contoured at +/- > 3 sigma, so you will always get some peaks, even with almost perfect models. > > > > Best, > > Herman > > > > *Von:*CCP4 bulletin board [mailto:[log in to unmask]] *Im Auftrag von > *Andrew Marshall > *Gesendet:* Mittwoch, 24. August 2016 08:31 > *An:* [log in to unmask] > *Betreff:* [ccp4bb] Disappearing difference density? > > > > Dear all, > > > > I have a 2.6 A structure at the validation stage, stats look good > (Rfree/Rwork : 25.7/19.7). I just have a patch of positive Fo-Fc density > present in a solvent-inaccessible hydrophobic pocket which I'm > struggling to fill. I've tried fitting glycerol (which seems unlikely, > but it's the only component of the crystallisation buffer that seems > about the right size) but the resulting Fo-Fc density is entirely > negative (no 2Fo-Fc density). When fitting a water molecule I get the > same result. > > > > I also tried the automated ligand identification tool in phenix which > suggested a number of possible small molecules (all placed in this patch > of difference density), including glycerol, but I get the same result > after a round of refinement. > > > > Has anyone observed anything similar before? > > > > Thanks, > > > Andrew Marshall > > PhD Candidate > > Laboratory of Protein Crystallography > > Dept. of Molecular and Cellular Biology > > School of Biological Sciences > The University of Adelaide > > >