From: Chen Zhao <c.z...@yale.edu>
Date: Thu, May 14, 2015 at 10:04 AM
Subject: Re: [ccp4bb] XDS Rmeas in space group determination
To: Kay Diederichs <kay.diederi...@uni-konstanz.de>
Hi Kay,
Thanks a lot for your reply! I am a little surprised to learn that the
centrosymmetry is always considered as a point groups symmetry component.
That might explain why all the anomalous data I have seen have higher Rmeas
than their native counterpart.
I looked at both points you suggested and I found that the high Rmeas is
because of radiation damage... Sorry about missing this point... But it is
good for me to learn how the Friedel pairs are treated as equal in point
group determination.
Thank you very much,
Chen
On Thu, May 14, 2015 at 3:18 AM, Kay Diederichs <
kay.diederi...@uni-konstanz.de> wrote:
> Hi Chen,
>
> if Rmeas is high (like 50 and up) even in P1 then maybe the integration
> was not right, or the indexing is offset by 1 in h or k or l ?
>
> To check the former, look at FRAME.cbf and see if the predictions match
> the spots.
>
> To test the latter try
> echo CENTRE | pointless XDS_ASCII.HKL
>
> best,
>
> Kay
>
> P.S. in XDS's space group determination, Friedels are indeed considered
> symmetry-related.
>
> On Wed, 13 May 2015 19:24:30 -0400, Chen Zhao <c.z...@yale.edu> wrote:
>
> >> Hi Ethan,
> >>
> >> Thanks a lot for your detailed information. I am aware that in IDXREF
> only the lattice symmetry was tried to be determined. I went back to check
> the subtrees in IDXREF because even for P1 the Rmeas is very high, meaning
> that the multiple measurements for the same reflections are already very
> imprecise (test resolution 10-5). I therefore am worried about multiple
> lattices.
> >>
> >> Also related to the probability thing you talked about, there is no
> point group has significantly low Rmeas in this case. Or it is just because
> even P1 has high Rmeas, so that the highest point group tried were
> considered to be correct? If so, it sounds hard to determine the point
> group in this case...
> >>
> >> Thank you so much,
> >> Chen
> >>
> >>
> >>>> On May 13, 2015, at 6:48 PM, Ethan A Merritt <
> merr...@u.washington.edu> wrote:
> >>>>
> >>>> On Wednesday, 13 May, 2015 18:17:04 Chen Zhao wrote:
> >>>> Hi Ethan,
> >>>
> >>> Sorry, I'm coming in late on this so I might have missed an
> >>> earlier explanation of exactly what programs are involved.
> >>>
> >>>
> >>>> Yes. My question was simply whether it calculates the statistics
> >>> ^^^^
> >>>> from completely unmerged intensities and just compare say h,k,l with
> -h,-k,l (or -h,-k,-l and h,k,-l) if there is a 2-fold? Although I believe
> so...
> >>>
> >>> What is "it"?
> >>>
> >>> If you mean the tables in IDXREF.LP, they only report the fit of points
> >>> to a particular lattice. They do not compare the intensities of
> >>> potential symmetry mates. Quoting from the program output:
> >>>
> >>> Note, that reflection integration is based only on orientation and
> metric
> >>> of the lattice. It does not require knowledge of the correct space
> group!
> >>> Thus, if no such information is provided by the user in XDS.INP,
> >>> reflections are integrated assuming a triclinic reduced cell lattice;
> >>> the space group is assigned automatically or by the user in the last
> >>> step (CORRECT) when integrated intensities are available.
> >>>
> >>> If you mean the output from a later run of pointless/aimless,
> >>> so far as I know it applies the symmetry operation being tested
> >>> to all reflections, which means that Friedel/Bijvoet pairs are
> >>> not compared. But I could be wrong on that point.
> >>>
> >>>> And what is a good number? Is 20 % OK? What about 30 % and even
> higher?
> >>>
> >>> Still refering to output from pointless/aimless, the crucial point is
> not
> >>> the absolute number but rather how the agreement for the symmetry
> operation
> >>> being tested compares to the agreement for the identity operation.
> >>>
> >>> For example, here is the output for a lousy data set with a real
> 2-fold:
> >>>
> >>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> >>> Scores for each symmetry element
> >>>
> >>> Nelmt Lklhd Z-cc CC N Rmeas Symmetry & operator (in
> Lattice Cell)
> >>>
> >>> 1 0.806 6.97 0.70 17852 0.516 identity
> >>> 2 0.919 7.67 0.77 21302 0.486 *** 2-fold k ( 0 1 0) {-h,k,-l}
> >>>
> >>> [snip]
> >>>
> >>> Laue Group Lklhd NetZc Zc+ Zc- CC CC- Rmeas R-
> Delta ReindexOperator
> >>>
> >>> 1 P 1 2/m 1 *** 0.919 7.30 7.30 0.00 0.73 0.00 0.50 0.00
> 0.1 [-h,-l,-k]
> >>> 2 P -1 0.081 -0.69 6.97 7.67 0.70 0.77 0.52 0.49
> 0.0 [h,-k,-l]
> >>> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
> >>>
> >>> In this case the program reports a 0.91 likelihood that the Laue
> >>> group is P2 even though the Rmeas is horrible.
> >>>
> >>> Ethan
> >>>
> >>>
> >>>> Thanks a lot,
> >>>> Chen
> >>>>
> >>>>
> >>>>>> On May 13, 2015, at 6:07 PM, Ethan A Merritt <
> merr...@u.washington.edu> wrote:
> >>>>>>
> >>>>>> On Wednesday, 13 May, 2015 17:51:59 Chen Zhao wrote:
> >>>>>> Hi all,
> >>>>>>
> >>>>>> I am sorry about this question which I should have figured out
> earlier. For
> >>>>>> point group determination, does the Rmeas consider Fridel pairs
> >>>>>> differently?
> >>>>>
> >>>>> A Friedel pair consists of the [hkl] and [-h-k-l] reflections.
> >>>>> This pairing is independent of space group.
> >>>>> So the agreement or lack of agreement between Friedel pairs is
> >>>>> not informative about selection of point group or space group.
> >>>>>
> >>>>> You may be thinking of a Bijvoet pair, which consists of
> >>>>> [hkl] and the Friedel mate of some symmetry equivalent of [hkl]
> >>>>> within a particular spacegroup.
> >>>>>
> >>>>> But even in the presence of anomalous scattering I think that
> >>>>> Bijvoet pairs are expected to agree with each other better than
> >>>>> with a reflection not related by point group symmetry.
> >>>>>
> >>>>>> (although I think it should be...) This is because I saw a
> >>>>>> derivative dataset collected at peak (from a demo) whose Rmeas is
> quite
> >>>>>> high (>50 %) for all the space groups tested (including P1).
> However, the
> >>>>>> native dataset has only <10 % Rmeas. Should I worry about the
> derivative
> >>>>>> dataset? There seems to be multiple lattices in both datasets based
> on
> >>>>>> IDXREF.
> >>>>>>
> >>>>>> You inputs are really appreciated!
> >>>>>>
> >>>>>> Sincerely,
> >>>>>> Chen
> >>> --
> >>> Ethan A Merritt
> >>> Biomolecular Structure Center, K-428 Health Sciences Bldg
> >>> MS 357742, University of Washington, Seattle 98195-7742
> >>>
>
