Here is a question about twinning:
I have a data set to 1.4 anstrom resolution. It can be reduced with
either
P61 or P6122 with almost equal Rsym. The protein is a homodimer with an
asymmetric organic ligand bound in a hydrophobic pocket within the dimer
and the ligand is not involved in crystal packing. The two fold of P6122
sits on the two fold of protein dimer. The asymmetric ligand sits on
the
two fold too.
I ran the twinning detection recommended by Dr. Yeates, here is the
result:
column 1: bin number
columns 2: upper resolution limit
columns 3: lower resolution limit
column 4: number of reflections in bin
column 5: average resolution in bin
column 6: <|I|^2>/(<|I|>)^2
column 7: (<|F|>)^2/<|F|^2>
column 8: fraction of theoretically complete data
<|I|^2>/(<|I|>)^2 is 2.0 for untwinned data, 1.5 for twinned data
(<|F|>)^2/<|F|^2> is 0.785 for untwinned data, 0.865 for twinned data
#bin | resolution range | #refl |
1 5.16 500.01 649 7.3161 2.0296 0.7591
0.8723
2 4.09 5.16 687 4.5452 1.8471 0.7988
0.9234
3 3.58 4.09 693 3.8107 1.9560 0.7790
0.9506
4 3.25 3.58 695 3.4035 2.0809 0.7767
0.9586
5 3.02 3.25 710 3.1279 1.9972 0.7828
0.9686
6 2.84 3.02 699 2.9231 2.1450 0.7798
0.9588
7 2.70 2.84 706 2.7646 1.9843 0.7665
0.9645
8 2.58 2.70 699 2.6355 2.2194 0.7664
0.9681
9 2.48 2.58 703 2.5281 2.0411 0.7782
0.9710
10 2.39 2.48 710 2.4354 2.2442 0.7571
0.9647
11 2.32 2.39 686 2.3561 2.0645 0.7796
0.9689
12 2.25 2.32 706 2.2856 2.0228 0.7932
0.9605
13 2.19 2.25 710 2.2223 2.0797 0.7710
0.9660
14 2.14 2.19 714 2.1652 2.0679 0.7773
0.9688
15 2.09 2.14 686 2.1145 1.9542 0.7895
0.9717
16 2.05 2.09 715 2.0680 1.8889 0.8073
0.9649
17 2.01 2.05 685 2.0254 2.0624 0.7974
0.9689
18 1.97 2.01 697 1.9870 2.0585 0.7818
0.9748
19 1.93 1.97 711 1.9502 1.9325 0.8013
0.9608
20 1.90 1.93 687 1.9162 2.1114 0.7999
0.9662
21 1.87 1.90 706 1.8846 2.1716 0.7985
0.9645
22 1.84 1.87 683 1.8550 2.0755 0.7979
0.9674
23 1.81 1.84 699 1.8272 2.1323 0.8067
0.9472
24 1.79 1.81 673 1.8010 1.8974 0.8240
0.9506
25 1.76 1.79 694 1.7763 1.7951 0.8334
0.9468
26 1.74 1.76 691 1.7524 1.7756 0.8450
0.9505
27 1.72 1.74 695 1.7299 1.7615 0.8449
0.9534
28 1.70 1.72 652 1.7090 1.6857 0.8568
0.9368
29 1.68 1.70 717 1.6886 1.6936 0.8549
0.9459
30 1.66 1.68 672 1.6691 1.8378 0.8468
0.9532
31 1.64 1.66 688 1.6506 1.8129 0.8549
0.9386
32 1.62 1.64 660 1.6334 1.6833 0.8627
0.9348
33 1.61 1.62 694 1.6163 1.5858 0.8838
0.9481
34 1.59 1.61 675 1.5999 1.7267 0.8674
0.9298
35 1.58 1.59 685 1.5843 1.6387 0.8779
0.9320
36 1.56 1.58 640 1.5690 1.9736 0.8602
0.9091
37 1.55 1.56 650 1.5552 1.5559 0.8955
0.9168
38 1.53 1.55 663 1.5412 1.5400 0.8896
0.9057
39 1.52 1.53 647 1.5274 1.5869 0.8876
0.8949
40 1.51 1.52 668 1.5146 1.4657 0.9021
0.9201
41 1.50 1.51 649 1.5019 1.4614 0.9161
0.9167
42 1.48 1.50 652 1.4899 1.3811 0.9203
0.8895
43 1.47 1.48 658 1.4780 1.3856 0.9219
0.9063
44 1.46 1.47 618 1.4667 1.3304 0.9279
0.8791
45 1.45 1.46 632 1.4555 1.2985 0.9319
0.8495
46 1.44 1.45 617 1.4447 1.2970 0.9297
0.8558
47 1.43 1.44 631 1.4344 1.2885 0.9321
0.8776
48 1.42 1.43 634 1.4240 1.2667 0.9348
0.8757
49 1.41 1.42 628 1.4143 1.2794 0.9317
0.8544
50 1.40 1.41 592 1.4048 1.4485 0.9257
0.8445
---------------------------averages-over-all-bins-----------------------------
<|I|^2>/(<|I|>)^2 = 1.8023 (2.0 for untwinned, 1.5 for
twinned)
(<|F|>)^2/<|F|^2> = 0.8381 (0.785 for untwinned, 0.865 for
twinned)
-------------------------------------------------------------------------
I refined the structure using P6122 with a protein monomer and ligand
in half occupancy using cns 0.5 R=22% and Rf=23%. Then generate
another monomer and put the model in P61 with the twinning law and a
fraction = 0.5, which give the same model and R = 20% and Rf=21%. Using
cns 0.9, anneal_twin.inp, R drop to17% but Rf increases to 25% and the
map genreated with model_map_twin is very bad in the area surrounding
the ligand. My question is:
Is this a twining or not? Which is a better refinement scheme: P6122
with monomer and half occupancy for the ligand or a dimer with the
twinning consideration? Thanks for help
Jun
Jun
