 |
 |
Hi Gihan, and anyone else who finds this interesting. Sorry for the late response: > how should one determine a point group and the space group of an > unknown crystal? > > I have a protein crystal with know unit-cell parameters. (these are > XFEL data so indexing wouldn't give the point and space groups). I > checked the PDB, but no luck the PDB structures have the different > space group assigned, no definitive answer hopefully, somebody can > point me in the right direction Space group determination using serial crystallography data is different to rotation crystallography because you have to start from the highest possible symmetry and work downwards by finding and resolving ambiguities, instead of merging in the lowest possible symmetry and working upwards by looking for possible symmetries. Just as with any data, the golden rule is that the space group is only a hypothesis until the structure is solved (and even then...). Here's a very brief step by step guide. Start by determining the cell parameters, which you've done already. Say the cell parameters look like a hexagonal P lattice. Proceed for a while on the assumption that it really is hexagonal P, but keep the golden rule in mind. In this case it might be, amongst others, monoclinic with two axes similar in length and an angle close to 120 degrees. Use your crystallographic knowledge to spot centering possibilities, for example a cubic F lattice might look rhombohedral with angles of 60 degrees (however, the indexing program should spot these for you). Merge the snapshots according to the highest point symmetry permissible by the lattice. You can look this up in many places including the symmetry chart distributed with CrystFEL: https://www.desy.de/~twhite/crystfel/twin-calculator.pdf It's always the point group with a grey background in the bottom left corner of the individual table for the lattice type. For the example hexagonal P lattice, it's point group 622. Do the standard tests on the data, particularly twinning tests including an L-test. Whenever you see apparently twinned data with serial crystallography, either the crystals are physically twinned or the true symmetry is lower and you need to resolve an indexing ambiguity and merge again. You will need to try rounds of ambiguity resolution until the tests are clean and the structure can be solved. The difficult part is finding your way through the maze of possible symmetries. You can attempt a resolution into any subgroup of the current symmetry, provided that the "ambiguity operator" is just a rotation (no reflections/inversions). The CrystFEL table shows the most obvious subgroups (the ones with the same lattice type). The comprehensive map of possibilities can be found in International Tables A, Fig 10.1.3.2 (in 5th edition). There are many special cases, and the find_ambi tool from the CrystFEL extra programs repository can help you find cases of "accidental" ambiguities due to the particular values of the lattice parameters: https://www.desy.de/~twhite/crystfel/programs.html If there are signs of twinning, try resolving the ambiguity into each of the subgroups. If the ambiguity resolves nicely (one way to tell is by the correlation coefficient graph, which should separate nicely: http://journals.iucr.org/j/issues/2016/02/00/zd5001/zd5001fig3.html), try merging the reindexed patterns in the lower symmetry point group, and check the twinning tests again. Once you have a merged set of reflections with no apparent twinning, things are the same as rotation crystallography. Examine the systematic absences to see if they suggest any screw axes. Try molecular replacement in all the possible space groups. Consider revisiting the earlier steps if there are problems. It's not easy, but it's also not that difficult, just different to usual. I would almost go as far as saying it's fun, like cracking a secret code. This kind of situation is my personal favourite part of crystallography (!) This recent paper embodies a similar workflow in a nice algorithm: https://journals.iucr.org/d/issues/2018/05/00/rr5155/ Hope that helps, Tom -- Thomas White <[log in to unmask]> <[log in to unmask]> 4E1F C14D 0E0A A014 FE5D 3FC6 C628 75D1 D4CA 4C30 Direct telephone: +49 (0)40 8998-5786