Hi, occupancy refinement in phenix.refine does not use information about special positions. So if it sees an atom (or groups of atoms) having partial occupancy it will refine them to be any number between 0 and 1. This is in my to-do list to fix one day. Currently, you need to turn off the occupancy refinement for these atoms manually using "occupancies.remove_selection". However, if the input occupancy is set to 1 then internally phenix.refine will do the right thing with respect to special position (it will apply the multiplicity factor), and it will not refine it. The downside of doing it this way is that this is not what PDB expects. Pavel. On 4/30/11 10:07 PM, Ralf W. Grosse-Kunstleve wrote:

This should work as you expect in more recent phenix versions. My recommendation is to install 1.7.1 and then try again. -- phenix versions can co-exist. Just make sure you source the phenix_env of the version you want to use. Ralf

------------------------------------------------------------------------ *From:* J. Fleming *To:* [email protected] *Sent:* Saturday, April 30, 2011 9:13 PM *Subject:* Re: [phenixbb] Sulfate ion on 2-fold axis

Hi all,

I have a sulfate ion sitting on a 2-fold axis of symmetry. If I set all the atom occupancies to 0.5 and refine using phenix.refine (Phenix v. 1.5-2) the occupancies are changed to 0.00 for the sulfur, 1.00 for three of the oxygen's, and 0.72 for the fourth oxygen. Can anyone explain this?

Thanks in advance, Jon

Hi, isn't it (or a simple derivative of) what phenix.refine outputs all the time, see log file for lines like this: | x-ray target function (ml) for work reflections: 4.379227 | where ML target function is defined here: Acta Cryst. (1995). A51, 880-887 Acta Cryst. (2002). A58, 270-282. I guess the above value is normalized by a number of reflections or something like this (I don't remember, but if you are interested let me know and I will have a look at the code). It may also not contain a constant part of the target. May be I should add it to phenix.model_vs_data output... Pavel.

Hi,

Is there any way in phenix (such as phenix.refine) to evaluate the LLG(log-likelihood) value by given the structure factor information only (Fo, sigmaFo and Fc)? Thanks in advance for any information!

Best regards, Hailiang

Hi Hailiang,

Yes, I understand it is part of phenix.refine output, but I am wondering whether there is a way to evaluate LLG by only using user-provided structure factor (Fo, sigFo, Fc).

internally in PHENIX there are many pieces of code that do this (like X-ray target calculation in phenix.refine), but I am not aware of any dedicated program that is exposed to the end-user and that does this. So I guess the quick answer is "no". If you think it is important to have then, as I mentioned before, I can and add it to phenix.model_vs_data output.

I think phenix.refine will have to read a PDB file, based on which model Fc will be calculated and used to evaluate LLG (instead of my provided Fc).

This is true. It will have to read a PDB file to obtain Fmodel (which is not just Fc).

I know there is a well-established LLG formula and may write a short code but I need to minimize sigmaA I think which take some efforts:-(

Not sure the code will be short since you have to estimate sigmaA (or alpha/beta - what's used on phenix.refine). By the way, in this case your input will also have to include free-R flags in addition to Fo and Fmodel. I recall I spent more than a week before I got it right and fast -;) Pavel.