Hi Randy,Many thanks for your explanation about the probability of twinning reported in PHASER for this dataset.In fact I was getting rather confused by the output from other programs like:AIMLESS
Estimated twin fraction alpha from cumulative N(|L|) plot 0.062 (+/-0.007)< |L| >: 0.457 (0.5 untwinned, 0.375 perfect twin)
Estimated twin fraction alpha from < |L| > 0.063
< L^2 >: 0.281 (0.333 untwinned, 0.2 perfect twin)
Estimated twin fraction alpha from < L^2 > 0.065
The L-test suggests that the data are not twinned
and REFMAC5**** Filtering out small twin domains, step 1 ****
Twin operators with Rmerge > 0.439999998 will be removedSymmetry operator L, -K, H : R_merge =0.422: twin is probable===> Warning: Problem with negative B--------------------------------------------------------------------------------------------------------------
**** Filtering out small twin domains, step 2 ****
Twin domains with fraction < 7.00000003E-02 are removed**** Twin operators with estimated twin fractions ****Twin operator: H, K, L: Fraction = 0.607; Equivalent operators: -H, K, -LTwin operator: L, -K, H: Fraction = 0.393; Equivalent operators: -L, -K, -HI guess that this dataset does not have a twinning problem.Best regards.Ray Brown
From: Randy Read <[log in to unmask]>
To: [log in to unmask]
Cc: [log in to unmask]
Sent: Friday, April 15, 2016 3:15 AM
Subject: Re: [ccp4bb] twin fraction
Dear Ray,
There is a formula that turns the second intensity moment into an estimated twin fraction, but it assumes that what you have is merohedral twinning, that the effect of measurement error is insignificant (which we try to assure in this test by only taking data from stronger resolution ranges) and that the systematic effects of anisotropy and tNCS have been removed successfully. We decided not to report that number, but rather some p-values that will suggest whether it would be a good idea to look at a more exhaustive set of twinning tests, like the ones in aimless or xtriage. I like the second p-value because the effect of a twinning fraction of less than 5% is not terribly significant. In your case, Phaser estimates that there’s about a 40% chance that the twinning fraction is less than 5%. The standard deviation on the second moment is quite large, so presumably you have a relatively small cell and relatively few reflections?
Anyway, the deviation from the expected value of 2 is not quite in the 5% significance range. I probably wouldn’t worry too much, but the programs that run a series of tests including the ones that take potential twin laws into account will give you a clearer picture.
We put this into Phaser for two reasons. One is that, having implemented a statistical analysis of the effect of tNCS, we were in a position to correct for a factor that other programs don’t handle as yet. The second is that, when people don’t yet realise that twinning has led them to merge their data in too high symmetry, they’re often stuck with unsuccessful molecular replacement, and a hint that twinning might be present should be useful at that point.
Best wishes,
Randy Read
-----
Randy J. Read
Department of Haematology, University of Cambridge
Cambridge Institute for Medical Research Tel: +44 1223 336500
Wellcome Trust/MRC Building Fax: +44 1223 336827
Hills Road E-mail: [log in to unmask]
Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk
> On 14 Apr 2016, at 18:05, Ray Brown <[log in to unmask]> wrote:
>
> Hi folks,
>
> I saw this twin information in the output from PHASER. Please could someone explain to me what exactly this means?
>
> What is the actual twin fraction for this dataset?
>
> Many thanks,
>
> Ray Brown
>
>
>
> tNCS/Twin Detection Table
> -------------------------
> No NCS translation vector
>
> -Second Moments- --P-values--
> Centric Acentric untwinned twin frac <5%
> Theoretical for untwinned 3.00 2.00
> Theoretical for perfect twin 2.00 1.50
> Initial (data as input) 2.30 1.89+/-0.082 0.08 0.404
> After Anisotropy Correction 2.31 1.88+/-0.081 0.0707 0.381
>
> Resolution for Twin Analysis (85% I/SIGI > 3): 3.62A (HiRes=2.81A)
>