Re: [phenixbb] Autosol help: FOM is great but map is trash

Hi Alex, I'm not a Phenix developer, but I have seen many cases like this, where reflections can be indexed with different cells, and one unit cell axis differs by a factor of 2. This just means that every second reflection (namely the odd one) along the reciprocal space direction corresponding to that axis is weak. It is not a priori clear which of the two choices is the best for structure solution and refinement. This depends on whether the weak reflections actually carry structural information (e.g. about tNCS with a translation close to 0.5), or whether they are a result of a crystal defect. It may sometimes happen that the crystal form is not suitable at all, since both choices don't refine well! Concerning P212121, P21212, P2122, etc: these are all primitive orthorhombic, and data processing alone cannot tell you which one is correct - they give identical data processing statistics. Typically, absent reflections along h00 0k0 00l should tell about screw axes, but in case of tNCS and two different cell choices the correct choice may not be obvious. Molecular replacement and subsequent refinement of all possibilities should help. Furthermore, contrary to what you write, it _is_ possible for all the orthorhombic space groups to account for reflections in both cell choices, so don't jump to conclusions too early. Just explicitly specify the (big or small) cell parameters in the data processing program of your choice! It may also help to try a different data processing package. Hope this helps, Kay Am 19.02.2021 um 04:28 schrieb [email protected]:

Message: 4 Date: Thu, 18 Feb 2021 18:15:15 -0500 From: Alex Johnson To:[email protected] Subject: [phenixbb] Autosol help: FOM is great but map is trash Message-ID:<[email protected]> Content-Type: text/plain; charset="utf-8"

Hi Phenix team,

I am attempting to phase a protein structure from SeMet-labeled crystals and have encountered some issues in using Autosol that appear unusual, as they have not been encountered by any members of my group. The short summary is that while I achieved great statistics from beam line collection and Autosol produces solutions with FOM (>0.4) from multiple crystals processed to ~2.4-3.0 A, the output maps are of poor quality and Autobuild can never build with Rfree <0.5. There is only one methionine for the ~30 kDa protein, but the crystals appear to have good anomalous signal with anomalous multiplicity >15 overall and delanon in the inner shell >0.5. There appears to be translational NCS and the dataset can be processed with good statistics in several space groups (P212121, P21212, P2122, etc) but this doesn?t solve the phase problem. The axis drops by half in processing in the P212121 versus P21212 space groups: from one crystal they are 48.37 101.56 110.39 90 90 90 (in P212121) and 101 .62 48.35 55.19 90 90 90 (in P21212).

If anyone in the community has encountered similar problems please let me know! I am all ears to any possible solutions and am also installing additional methionine for labeling but unsure whether this will overcome the other issues.

Thanks,

Alex Johnson Postdoc, Kranzusch lab, Dana-Farber Cancer Institute [email protected] mailto:[email protected]