Hi Magnus,

In one of such similar cases, what I found useful was to reprocess the data in P1 (not just expanding..), run MR-if successful then initial cycles of rigid body refinement..and then feed the P1 model and mtz file to zauda...it should tell you (at least in my case it did..!!!) the correct space group. Again one word of caution is, sometimes the tNCS could hide twining, so you need to take care of that also.

Another thing, do not be rigid about the space group what you got from pointless/SCALA...when you are doing the MR ...be flexible to check both P2 and P21...(presence of tNCS mimicks the systematic absence conditions).

Good luck..!!!

Sudipta.

On Fri, May 13, 2016 at 12:09 PM, Ethan A Merritt <[log in to unmask]> wrote:

On Friday, 13 May, 2016 16:14:47 Magnus Alphey wrote:
> Dear CCP4bb,
> I am having a little issue with some data that on the surface looks OK but it is not playing ball.
> The target is a hetero-octamer - 4 each of 2 subunits.
> Autoindexing with iMosflm appears to nail the lattice as mP (Penalty of 1; next best is mC with penalty of 40)

Penalty during auto-index is one thing, success in processing is another.
Did you try integration+scaling with the mC setting?

        Ethan

> Cell dimensions are  94, 147,102, 90, 102, 90.
>
> Scala gives
>
>                                            Overall  InnerShell  OuterShell
>   Low resolution limit                       31.47     31.47      2.64
>   High resolution limit                       2.50      7.91      2.50
>
>   Rmerge                                     0.120     0.039     0.710
>   Rmerge in top intensity bin                0.041        -         -
>   Rmeas (within I+/I-)                       0.144     0.047     0.859
>   Rmeas (all I+ & I-)                        0.144     0.047     0.859
>   Rpim (within I+/I-)                        0.080     0.026     0.478
>   Rpim (all I+ & I-)                         0.080     0.026     0.478
>   Fractional partial bias                   -0.033    -0.012    -0.084
>   Total number of observations              273802      9515     34527
>   Total number unique                        90106      2968     11962
>   Mean((I)/sd(I))                              6.8      17.6       1.6
>   Mn(I) half-set correlation CC(1/2)         0.993     0.998     0.557
>   Completeness                                96.8      98.3      88.6
>   Multiplicity                                 3.0       3.2       2.9
>
>   Anomalous completeness                      88.5      97.2      73.5
>   Anomalous multiplicity                       1.6       1.7       1.6
>   DelAnom correlation between half-sets     -0.018     0.111     0.003
>   Mid-Slope of Anom Normal Probability       0.959       -         -
>
> Outlier rejection and statistics assume that there is no anomalous scattering
>
> Average unit cell:    94.03  146.71  101.90       90.00  102.41   90.00
>
> Space group: P21
>
> Average mosaicity:     0.83
>
> Minimum and maximum SD correction factors: Partials   0.65 155.76
>
>
> But then cTruncate gives
>
> TRANSLATIONAL NCS:
>
> Translational NCS has been detected at ( 0.000,  0.500,  0.500).
> The probability based on peak ratio is  0.02% that this is by chance (with resolution limited to  4.00 A).
> This will have a major impact on the twinning estimates and effectiveness of the truncate procedure
>
> Peak#       Location       Ratio  Q-score
>  1    ( 0.00, 0.50, 0.50)  0.44   0.02
> The analysis uses the peak heights in the patterson map that are further than 14 A (approx. 4 Ca-Ca) from the origin.  The presence of a large off origin peak (above 20%) and/or a very low Q-score, below 1.0, is a string indicator of the presence of tNCS.  An intermidiate Q-score, between 5.0 and 1.0, may indicate weak tNCS or be the result of cross vector of a large scatterer such as a cluster or heavy metal.
>
>
> Crystals from similar drops have previously autoindexed as C2
> 101.8, 146.7, 94.9, 90, 102.5, 90
>
> Molecular replacement has not been successful (so far) and I am wondering if the translational NCS is causing an issue. Perhaps the space group/cell dimensions are not as autoindexing suggests if the tNCS interfering. Should I be treating the data some specific way ?
> If anyone has experience of this and can provide advice/suggestions, then that would be much appreciated.
>
> Kind regards,
>
> Magnus
>
>
>
>

--
Ethan A Merritt
Biomolecular Structure Center,  K-428 Health Sciences Bldg
MS 357742,   University of Washington, Seattle 98195-7742