# [phenixbb] pseudo-merohedral twinning in P212121

Peter Zwart PHZwart at lbl.gov
Wed Sep 12 08:19:54 PDT 2007

```Sometimes low data quality can give L-tests statistics that are fishy
and make it look twinned, even while the data is bad.

Moving down the ladder of spacegroups to make sure that you have a
solution is I think dangerous, unless you have an extremely reason to
do so.Not being able to solve your structure is a good reason, but
this step should not be taken lightly.

The fact that you have less clashes in P212121 might be due to the
fact that you have 'more room' in that spacegroup.

Did you try

all possible group that fall under P4x2x2 and P4x? Those would really
be the first things to try.  Note that P212121 is both a subgroup of
P41212 and P43212:
P 21 21 21 (a+1/4,b,c-3/8)  --->  P 41 21 2  ::    using: (-y+1/2,x+1/2,z+1/4)
P 21 21 21 (a+1/4,b,c-1/8)  --->  P 43 21 2  ::    using: (-y+1/2,x+1/2,z+3/4)

Obtaining better experimental phases might do the trick. Iyou do want
to go the sub spacegroup route,  these are subgroups for P43212:
P 21 21 21 (a+1/4,b,c-1/8)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)    symops left: 0
C 1 2 1 (x+y,-x+y,z)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)  (y+1/2,-x+1/2,z+1/4)  (-x,-y,z+1/2)    symops
left: 0
C 1 2 1 (x-y,x+y,z-1/4)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)  (y+1/2,-x+1/2,z+1/4)  (-x,-y,z+1/2)    symops
left: 0
P 1 21 1 (b-1/8,c,a-1/4)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)  (y+1/2,-x+1/2,z+1/4)  (-x+1/2,y+1/2,-z+3/4)
symops left: 0
P 1 21 1 (a-1/4,b,c-3/8)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)  (y+1/2,-x+1/2,z+1/4)  (x+1/2,-y+1/2,-z+1/4)
symops left: 0
C 2 2 21 (x-y,x+y,z)  --->  P 43 21 2  ::    using:
(-y+1/2,x+1/2,z+3/4)    symops left: 0
P 43 (a,b+1/2,c)  --->  P 43 21 2  ::    using: (x+1/2,-y+1/2,-z+1/4)
symops left: 0

the operators listed as 'using' would then be broken symops whose
rotational part is a possible twin law.

HTH

P

2007/9/12, Schneider Sabine <paxss2 at nottingham.ac.uk>:
>
>
>
>
> Hi everyone,
>
>
>
> I have a data set to 2.8A, with most likely 2 molecules in the asu and a
> molecular replacement model with 40% sequence identity. (Cell 71.3 71.3
> 227.9 90 90 90)
>
>
>
> Initially the data appeared to be either P4322 or P43212 (Rsym = 7.4%; Cell
> 71.3 71.3 227.9 90 90 90), but the molecular replacement solutions had major
> clashes with the symmetry related molecules and a lower LLG for the solution
> of the second molecule. No twinning was detected by Xtriage.
>
>
>
> Processing the data in P212121 (Rsym 5.6%, cell 71.3 71.4 228.2 90 90 90)
> gave me a better molecular replacement solution (LLG 121/373), but I had 5
> clashes with the second molecule in the asu. (I tested all alternative
> spacegroups)
>
>
>
> I also have a MAD dataset (very low quality!) and I combined the phases with
> the one obtained from molecular replacement in Sharp. After NCS averaging
> the maps had improved (still not great). I tried to improve things in the
> model and run Phaser again and the solution had better statistics (LLG
> 338/1223) and the clashes were reduced to 2.
>
> Xtriage detected pseudo-merohedral twinning with a twin fraction of 43% with
> twin operator –k, -h, -l.
>
> So I tried refining the model with phenix telling it the twinning operator
> and fraction, but R/Rfree are stuck around 45%. I also tried simulated
> annealing but that gave even higher R/Rfree.
>
>
>
> Is there anything I can do, or is it a hopeless case?
>
>
>
>
>
>
> Sabine
>
>
>  This message has been checked for viruses but the contents of an attachment
> may still contain software viruses, which could damage your computer system:
> you are advised to perform your own checks. Email communications with the
> University of Nottingham may be monitored as permitted by UK legislation.
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://www.phenix-online.org/mailman/listinfo/phenixbb
>
>

```