Dear Mohammed,
As has been mentioned before, the first thing to do is to check your data, is it twinned? I would also have a look at the images: are there strong ice rings, is the data very anisotropic?
If the data looks ok, the next suspect is the space group. Always let your molecular replacement program test all possible space groups! In case you tested only C2221, Phaser will place the search molecules in C2221 and Zanuda will confirm that molecular replacement had been run in C2221. Zanuda will find overlooked cases of higher symmetry, but not incorrectly assigned symmetry, or cases where the symmetry is lower. To determine the space group with Zanuda, you might want to process und run MR in a lower symmetry space group, e.g. P1 or C2.
Your diffraction data is a convolution of the molecular transform of your protein molecules (alone, without crystal packing) and the transform of the crystal packing. This means that if the orientation of your protein molecules is correct, but if you misidentified e.g. one screw axis, you may already get quite good Rfactors (35-40%) and electron density, and you may even see bound ligands. However, the Rfactor is stuck. You may even have non-crystallographic symmetry, masquerading as crystallographic symmetry, so I would try processing your data in various lower symmetry space groups and run MR in those space groups, testing all possibilities!
Best regards,
Herman
-----Ursprüngliche Nachricht-----
Von: CCP4 bulletin board [mailto:[log in to unmask]] Im Auftrag von Mohamed Noor
Gesendet: Montag, 27. Juni 2016 02:08
An: [log in to unmask]
Betreff: [ccp4bb] Rfree stuck at around 39 %
Dear all
I am working on a structure that extends to about 2.5 A (CC* of 0.5) in the SG C 2221 (a=134.48, b=424.9, c=84.49). The long b axis is probably why the solvent content is 75 % with 2 NCS copies. The images were processed with XDS in Xia2.
After building and refining the model (Phenix+Refmac and one round of PDB_REDO), the Rwork/Rfree is about 34/39 %. Even the addition of water molecules does not change this, although only less than 10 molecules can be placed.
I ran Zanuda from the old GUI on my data and model - it suggests that the original SG is the correct one. I can easily see the density for my naturally-bound ligands (heme and iron-sulfur cluster), which makes me think the solution from Phaser run is indeed (almost) correct. The model statistics are also fairly good - Molprobity clashscore 37, RMSD angles 1.39, RMSD bonds 0.007, average B 90.4. Some loops have poor density, so I built them manually after superposing the NCS copies.
Is there a list of things that I should check to identify the source of the problem?
Thanks.
Mohamed
|