Thank you, Manfred. This is a very important point to remember. On Wed, Apr 18, 2018 at 9:13 PM, Gihan Ketawala <gketa...@asu.edu> wrote:
> Hi, > my question is; > 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 > > Best, > Gihan > Daniel M. Himmel, Ph. D. wrote: Gihan, The best way, in my opinion, to choose a suitable (and correct) space group is to combine your own understanding and knowledge with the available software tools. I’ve never indexed XFEL data, so I don’t know if this data can be sampled and run through hkl2000/3000 or Mosflm to get some first guesses. However, scaling the data in different space groups can be helpful (such as if you just have to decide which screw axis, if any, is present) to see which one gives the best statistics and best systematic absences results. In any case, I would run the CCP4 program “pointless” to get some space group suggestions. Then, think about what systematic absences (if any) are expected for space group candidates, and look at slices through the reflections data in reciprocal space (such as using CCP4 hklview) to see if you can approximately recognize the expected systematic absences. Now, calculate the Matthews coefficient (Vm) on likely space groups. Each space group has its own number of asymmetric units per unit cell. You will need to play with the number of protein molecules per asymmetric unit to get the most reasonable Vm. That will also give you an important prediction to compare to what you already know about the protein and to confirm once you solve the structure and check the electron density map. I would encourage you to calculate Vm manually, in addition to computing it in CCP4 or your favorite other crystallographic package. Conventionally, the ideal Vm for proteins is said to be 2.7, but the actual acceptable range of values depends on the assumed protein crystal density (related to water content) and the resolution of the data [see Kantardjieff & Rupp. Protein Science 12:1865-1871 (2003) for a useful guide]. You also want the highest symmetry that is consistent with your data. If you still haven’t narrowed the space group down to one choice (or when there are space group ambiguities), and you have reasonable search models or homology structures to try, run ccp4 phaser with each of the space group choices with each of the search models and choose the space group that gives you the best result. Alternatively, or in addition, if you have anomalous data, you can solve phases by SAD in each of the likely space groups. As a final check, view the electron density of the chosen space group solution to be sure you see continuous peptide chains. I’m sure there are other shortcuts I have not mentioned. People with XFEL experience can chime in with advice. Whichever software tools you use to help you get the space group, don’t treat the software like a black box. In the end, everything you know in advance about the protein and the results you are getting should make sense (for instance, whether it’s a monomer, trimer, etc.), unless you discovered something completely novel. Many times, getting the space group is quick and straightforward. Sometimes, it takes a lot of work and thinking. -Daniel Weiss, Manfred wrote: Daniel, Gihan and all, please don't forget that the correct space group is ONLY determined once youhave refined the structure. Everything before is just an assumption. And there is many things that can fool you. Just to mention twinning, tNCS, etc. Therefore, when it comes to data collection from unknown crystals, I strongly recommend to play it safe and always collect at least 180 degrees of data, prefereably more. Best Manfred
