[phenixbb] General Questions about Joint x-ray/neutron refinement

Pavel Afonine pafonine at lbl.gov
Fri Jun 17 01:53:05 PDT 2011


Hi Louis,

I guess you are talking about the data that you sent me a few weeks 
ago... Let's see.

X-ray data:
- resolution ~1A,
- completeness ~97%,
- starting R-factors ~17/18%.

Neutron data:
- resolution ~2A,
- completeness: overall ~83%, in bins:
Bin  Resolution Range  Compl.
    1 28.2899 -  5.9817    0.98
    2  5.9817 -  4.7552    0.98
    3  4.7552 -  4.1563    0.96
    4  4.1563 -  3.7772    0.94
    5  3.7772 -  3.5070    0.94
    6  3.5070 -  3.3006    0.91
    7  3.3006 -  3.1355    0.89
    8  3.1355 -  2.9992    0.89
    9  2.9992 -  2.8838    0.87
   10  2.8838 -  2.7844    0.87
   11  2.7844 -  2.6974    0.83
   12  2.6974 -  2.6204    0.82
   13  2.6204 -  2.5514    0.81
   14  2.5514 -  2.4892    0.80
   15  2.4892 -  2.4327    0.79
   16  2.4327 -  2.3809    0.76
   17  2.3809 -  2.3333    0.78
   18  2.3333 -  2.2893    0.76
   19  2.2893 -  2.2484    0.77
   20  2.2484 -  2.2103    0.74
   21  2.2103 -  2.1747    0.73
   22  2.1747 -  2.1412    0.67
   23  2.1412 -  2.1098    0.74
   24  2.1098 -  2.0801    0.69
   25  2.0801 -  2.0520    0.69
   26  2.0520 -  2.0253    0.65
   27  2.0253 -  2.0000    0.66
so, essentially it is (most optimistically) ~2.5A resolution or so.
- starting R-factors ~35/37%

Now, let's do the best refinement we can against X-ray and neutron data 
(not joint, separately). I just did, and here is what I'm getting:

X-ray: r_work = 0.1229 r_free = 0.1357 (starting: 17/18%)
Neutron: r_work = 0.2791 r_free = 0.3489 (starting: 35/37%)

My conclusion from this is: the X-ray model looks so good so the poor 
neutron one is unlikely to bring any new useful information in this 
particular case. Given this specific combination of X-ray and neutron 
data qualities, I'm not sure the joint XN refinement makes sense at all 
(at least the way it is curently implemented in phenix.refine). 
Combining these two data sets in joint XN refinement will probably 
worsen X-ray derived structure and may be slightly improves Neutron one, 
but the question is: what is the new information that you are hoping to 
obtain by doing this? Once again, your very good 1A resolution X-ray 
model should answer all your questions, and it is unlikely that the poor 
neutron model derived from barely complete 2.5A resolution data with 
R-factors 28/35% will add any more of useful information.

May be collecting a better neutron dataset is a way to go (in case your 
very high resolution X-ray structure still doesn't answer the questions 
that prompted this structure solution project).

If you still want to experiment with this, I think a better option would 
be to obtain the best X-ray derived structure and use it as a reference 
model in refinement against the neutron data alone. See more about using 
reference model in refinement here:
http://www.phenix-online.org/documentation/refinement.htm#anch30

Coming back to your original question: I'm not sure it makes sense to 
compare water structures between structures obtained from such a 
different (in quality) data.

I'm open to further discussion. Let me know if you need any help with 
this or have any questions.

All the best!
Pavel.


On 6/16/11 1:07 PM, Louis Lazar wrote:
> Developers/all:
> I have been running some Joint X/N jobs, and I have some questions 
> that perhaps you can help me with. First, doing joint refinement gave 
> a total of 53 waters (thanks to ordered_solvent=true), yet the 
> electron density seems to show that there are several more waters that 
> are actually present (however it should be noted that there is no 
> nuclear density associated with those waters, and in some cases no 
> nuclear density associated with the 53 that were chosen by 
> ordered_solvent=true). Should I add in manually to the electron 
> density what I believe are these extra waters, or simply ignore them?
>
> Second, when I try to real space refine any of the residues, I get an 
> error message telling me about exploding atoms. What is the cause of 
> this message, and is it possible to real space refine any of these 
> residues and specifically the hydrogen atoms associated with them.
>
> Third, during the joint refinement run, I chose to refine occupancies 
> of all atoms. This breaks down into several questions: A) Should I be 
> refining occupancies at 2.0 A resolution (of the neutron structure; 
> 1.05A for the X-ray)? B) How does Phenix go about refining occupancies 
> of hydrogen/deuterium sites? C) If the resulting occupancies result in 
> values of, for example, 60% deuterium and 40% hydrogen, am I to 
> presume that deuterium presence is significant if the nuclear density 
> looks "good". At what point of deuterium occupation (while examining 
> the nuclear density), would I say that there is likely no deuterium 
> present (or that it is insignificant)?
>
> Hopefully all of these questions make sense, and if I need to clarify 
> anything, please let me know. Thanks so much in advance for the help, 
> I appreciate it as always.
>
> Regards,
>
> Louis



More information about the phenixbb mailing list