Re: [ccp4bb] Rigid body refinement as last refinement?
> I am annoyed at the fact that Refmac doesn't seem to have much respect > for the electron density map Other people are much better qualified than me to point this out, but maximum likelihood methods have the utmost respect for your _data_ (and especially their standard deviations). What you think you see in the map is not what your data really 'say'. -- Dr Nicholas M. Glykos, Department of Molecular Biology and Genetics, Democritus University of Thrace, University Campus, 68100 Alexandroupolis, Greece, Fax +302551030613 Tel ++302551030620 (77620), http://www.mbg.duth.gr/~glykos/
Re: [ccp4bb] Rigid body refinement as last refinement?
Hi, On Fri, Feb 06, 2009 at 08:13:29AM -0800, Andy Millston wrote: > Restrained refinement does a good job at refining most of the > structure except a small region with poor density. Every time I > refine it, it puts this region completely out of the map. Do you mean: a) after restrained refinement the 2Fo-Fc map tells me the model should be were it was before refinement (and were it stays during rigid body refinement) or b) after restrained refinement the 2Fo-Fc map and the model don't match up. If it's a), i.e. 2Fo-Fc map before and after restrained refinement is very similar but model has been moved away: are you sure there isn't some very bad contact that pushes your model away? Have you checked e.g in Coot with all symmetry-equivalents switched on? If it's b): maybe the initial 2fo-Fc map (after rigid-body refinement) was heavily biased and showed you an artefact? And the restrained refinement is actually telling you the truth - your model isn't were you thought it should be and there is no data/information to put it there. > It doesn't > have much effect on overall geometrical factors as its a very small > part of a relatively large molecule. > I am annoyed at the fact that Refmac doesn't seem to have much > respect for the electron density map while refining this region, But it does (well: it doesn't look at the density per se - for that you would need to do real-space refinement): just that something else (probably one of its geometry terms) it telling REFMAC to do what it does in the end. > and all my attempts to keep the residues within the map go in vain. Remember that the map isn't something fixed (unless you have very good experimental phases): it's based on the model phases after all. > The refined residues are still ok according to Fo-Fc map, but > they're completely out of the 2Fo-Fc map boundaries. What do you mean with '2Fo-Fc map boundaries'? > Also, my model and the refined model both look bad on Ramachandran > plot. That's ok, we're not refining against Ramchandran plot (or at least we shouldn't) - so as others have mentioned: this is telling you something. > But I think its a unique feature of the molecule, at least thats > what the density suggests. Also, packing score and quality value > plot suggest that there is nothing seriously wrong with my model. A lot of these criteria might not pick up details in small regions. Furthermore, if you start e.g. from a good initial model (molecular replacement) then all the purely geometrical quality indicators will be great, since nothing has changed there after rigid-body refinement. > As you've suggested I should try altering the restraints and see how it goes. Or combine restrained refinement for most of your protein with this part as a rigid body? I don't know if REFMAC can do that though ... Or leave that part out for the moment and concentrate on getting everything else as good as possible: this would give you the best phases frmo the 'good region' and maybe then the 2Fo-Fc and Fo-Fc map tell you what's going on in the other part. Sometimes its very beneficial to remove unclear regions again to not bias your own expectations of what it should look like. Cheers Clemens -- *** * Clemens Vonrhein, Ph.D. vonrhein AT GlobalPhasing DOT com * * Global Phasing Ltd. * Sheraton House, Castle Park * Cambridge CB3 0AX, UK *-- * BUSTER Development Group (http://www.globalphasing.com) ***
Re: [ccp4bb] Rigid body refinement as last refinement?
Hi Andy: Just be very careful of special pleading type arguments. I really think this is trying to tell you something, along with the Ramachandran plot. On a more practical level, the PDB will flag it when you try to deposit it. There is something wrong. It might be worth making a composite-omit map in CNS, and if you have experimental phases, go back and look at the original MIR (or whatever) map. Double-check the space group is right. Do all that before you change any defaults. It is very good by the way you are asking for feedback at this stage. If it turns out there is a problem, it is much better to catch it now. Also, if you do have to do something non-canonical during the refinement, the burden is placed on the paper's authors to justify why. So either way this will be time well spent. Best of luck. Bill On Feb 6, 2009, at 8:13 AM, Andy Millston wrote: Thank you for your responses. Restrained refinement does a good job at refining most of the structure except a small region with poor density. Every time I refine it, it puts this region completely out of the map. It doesn't have much effect on overall geometrical factors as its a very small part of a relatively large molecule. I am annoyed at the fact that Refmac doesn't seem to have much respect for the electron density map while refining this region, and all my attempts to keep the residues within the map go in vain. The refined residues are still ok according to Fo-Fc map, but they're completely out of the 2Fo-Fc map boundaries. Also, my model and the refined model both look bad on Ramachandran plot. But I think its a unique feature of the molecule, at least thats what the density suggests. Also, packing score and quality value plot suggest that there is nothing seriously wrong with my model. As you've suggested I should try altering the restraints and see how it goes. Thank you. Andy
Re: [ccp4bb] Rigid body refinement as last refinement?
Thank you for your responses. Restrained refinement does a good job at refining most of the structure except a small region with poor density. Every time I refine it, it puts this region completely out of the map. It doesn't have much effect on overall geometrical factors as its a very small part of a relatively large molecule. I am annoyed at the fact that Refmac doesn't seem to have much respect for the electron density map while refining this region, and all my attempts to keep the residues within the map go in vain. The refined residues are still ok according to Fo-Fc map, but they're completely out of the 2Fo-Fc map boundaries. Also, my model and the refined model both look bad on Ramachandran plot. But I think its a unique feature of the molecule, at least thats what the density suggests. Also, packing score and quality value plot suggest that there is nothing seriously wrong with my model. As you've suggested I should try altering the restraints and see how it goes. Thank you. Andy
Re: [ccp4bb] Rigid body refinement as last refinement?
On Feb 6, 2009, at 6:58 AM, Andy Millston wrote: I am told by a mentor that rigid body refinement should never be the last refinement before submission to pdb. Any idea why? It is counter-productive and he/she doesn't want to get jumped and beaten up in the parking lot at the next conference. Structures in which every round of retrained refinement messes up a conformation, is there any alternative to not using rigid body refinement? Find out why the conformation is being "messed up" because that is usually telling you something has been fit wrongly. Andy Bill
Re: [ccp4bb] Rigid body refinement as last refinement?
On Feb 6, 2009, at 15:58, Andy Millston wrote: I am told by a mentor that rigid body refinement should never be the last refinement before submission to pdb. Any idea why? When restrained refinement 'messes up a conformation' it is extremely likely it does it for a perfectly sound reason. This could be that for example that the conformation is wrong, or that the restraints used to describe it are wrong. You should use that as a warning that your structure needs more work. Structures in which every round of retrained refinement messes up a conformation, is there any alternative to not using rigid body refinement? Yes, work more on the refinement. Did you validate your structure using eg MolProbity? A. Andy P please don't print this e-mail unless you really need to Anastassis (Tassos) Perrakis, Principal Investigator / Staff Member Department of Biochemistry (B8) Netherlands Cancer Institute, Dept. B8, 1066 CX Amsterdam, The Netherlands Tel: +31 20 512 1951 Fax: +31 20 512 1954 Mobile / SMS: +31 6 28 597791
Re: [ccp4bb] Rigid body refinement as last refinement?
Andy, you should probably direct this question to the mentor first. What is her explanation? What do you mean by "refinement messes up a conformation"? Sometimes with poor quality data refinement might lead to poor Ramachandran map. You fix it manually, only to see it deteriorating again. You may try to play with the weight of geometry restraints. Since polypeptide backbone conformation is forced to occupy certain areas of (phi,psi) space because of steric clashes, one would expect that increasing contribution of van der Waals repulsion term may help. I heard an opinion (which I share to some degree) that it is never a good idea to enforce Ramachandran directly, by setting (phi,psi) restraints, but you may try, for instance, restraining the hydrogen bonds that stabilize the secondary structure. What this means ideologically is that your data is not of good enough quality, so you incorporate the knowledge of secondary structure. Think of it as incorporating additional experimental data to boost your data/parameter ratio. If I understand correctly, what you want to do is to fix the model manually and then just do rigid body to tidy things up a bit. Alternative could be to fix/restrain the problematic parts of the structure and do refine the rest. HTH, Ed. On Fri, 2009-02-06 at 06:58 -0800, Andy Millston wrote: > I am told by a mentor that rigid body refinement should never be the > last refinement before submission to pdb. Any idea why? Structures in > which every round of retrained refinement messes up a conformation, is > there any alternative to not using rigid body refinement? > > Andy > > -- Edwin Pozharski, PhD, Assistant Professor University of Maryland, Baltimore -- When the Way is forgotten duty and justice appear; Then knowledge and wisdom are born along with hypocrisy. When harmonious relationships dissolve then respect and devotion arise; When a nation falls to chaos then loyalty and patriotism are born. -- / Lao Tse /
