<div dir="ltr">On Sat, Sep 13, 2014 at 3:03 PM, Pavel Afonine <span dir="ltr"><<a href="mailto:pafonine@lbl.gov" target="_blank">pafonine@lbl.gov</a>></span> wrote:<br><div class="gmail_extra"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">
<div bgcolor="#FFFFFF" text="#000000">In the first case the bulk-solvent mask will be set in the ligand
region and therefore it will mask ligand density (bulk-solvent will
be filled into the ligand region). Depending on the strength of
ligand density it may be masked completely or deteriorated.<br></div></blockquote><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div bgcolor="#FFFFFF" text="#000000">
If you follow the second option you will always get positive density
in ligand area. This density may correspond to bulk-solvent, ligand
or mixture of both. That is there will be no simple way to
differentiate whether this density arises from the ligand or
bulk-solvent.<br></div></blockquote><div><br></div><div>I thought phenix.refine ignores zero-occupancy atoms when calculating the bulk solvent mask? In fact there is an option specifically to toggle this option, and I've noticed very different results with and without it.</div><div><br></div><div>Wei: It's fine to use the options you mentioned for this - however, weight optimization is very slow and not really necessary since you are only running SA to remove phase bias, not to get the best possible model. I usually just use individual_sites+individual_adp plus whatever extra restraints are appropriate (NCS, secondary structure, etc.). I wouldn't worry if you end up with an overfit model with worse geometry - as long as it has been sufficiently shaken (but not too much!) and the ligand difference density is clear, you are fine. Just be careful that in the process of running SA, you haven't actually introduced new bias into the maps - but the mFo-DFc map is relatively safe (the 2mFo-DFc map much less so).</div><div><br></div><div>Restraining the zero-occupancy ligand (and perhaps adjacent residues) is strongly recommended, especially if it contains heavier elements. (The newer versions of the phenix.refine GUI have a button in the Output tab to set this up.) Without restraints, you run the risk of refining the surrounding model into the ligand density. As always this is worse at lower resolution - at 2.8Å I'm not sure what to expect.<br></div><div><br></div><div>Perhaps one day we will write up something more formal about map "validation", since there seem to be a lot of misconceptions about these methods and they are not always used appropriately.</div><div><br></div><div>-Nat</div></div></div></div>