Pavel Afonine wrote:
I suspect that negative peaks are from the bulk-solvent mask, as you suspected.
The bulk solvent is flat... at least phenix.refine uses Flat Bulk solvent model (mask based). If bulk solvent is guilty for this, why would you observe peaks ?
I repeat once again: looking at Average Kick map may resolve this puzzle.
Pavel.
Sometimes there are voids just big enough for a tiny bit of solvent mask to fill. This results in negative difference density. Even though the mask is flat, the B-factor adjustment will convert small , isolated fragments of bulk mask into an approximately gaussian sphere, producing a well-defined negative peak. However, this situation is fairly rare. Voids big enough for the bulk mask to enter are usually big enough for a water. A true hydrophobic void deters from protein folding stability, and normally only occur if it pertains to the protein's function. So, a simple noise peak probably is more likely. The only way to know for sure is to make a bulk-solvent map. Is it possible to write out bulk solvent coefficients? Joe Krahn