Re: [ccp4bb] Refinement problem
Rather hard to advise! Options are: a) The helical domain is wrongly positioned! b) There are scaling problems or swathes of missing data which can depress some parts of the electron density. If you use coot and reduce the contour level in this region, is anything sensible visible? A useful test is to set the occupancy of some bulky residues to 0.00, repeat the refinement, then see whether you can see the missing atoms at some level in the map. Eleanor On 10 September 2014 03:20, Appu kumar wrote: > Hello All, > > I am new to low resolution refinement parametrization and regularization. > Crystal diffracted with high anisotropy reaching to 3.5A in one direction > and 4.5A other direction. I am refining a structure at 3.9A resolution. > Protein has two domain connected trhough a linker and is packed as tetramer > in ASU. Refinement in phenix as well as in refmac leads to wipe away of > electron density in helical domain of protein leaving only blobs of > electron density while other domain have good amount of density after > refinement. I need your precious and valuable suggestion for proceeding > with refinement. > > Thanks in advance for your suggestion. > > Thank you > >
[ccp4bb] Refinement problem
Hello All, I am new to low resolution refinement parametrization and regularization. Crystal diffracted with high anisotropy reaching to 3.5A in one direction and 4.5A other direction. I am refining a structure at 3.9A resolution. Protein has two domain connected trhough a linker and is packed as tetramer in ASU. Refinement in phenix as well as in refmac leads to wipe away of electron density in helical domain of protein leaving only blobs of electron density while other domain have good amount of density after refinement. I need your precious and valuable suggestion for proceeding with refinement. Thanks in advance for your suggestion. Thank you
Re: [ccp4bb] refinement problem
At this resolution it is very much possible to choose a wrong space group. So try lower symmetry first and see if it helps. Check for twinning too. On Sat, 2011-01-08 at 19:08 +, Dimitris Ladakis wrote: > Dear all > I've got a 3A dataset processed with mosflm and scaled. I've > runned MOLREP that gave a solution with a score of 0.52. At this > stage the Rfactor was 44%. > After my first round of refinement with REFMAC i've got both Rfactor > and an Rfree at about 43%. After model building and several refinement > rounds i've got an Rfactor 36% and the RFree is 45%. Since there are > two molecules in the AU i tried to do NCS phased refinement using > NCSREF, but it didn't do anything apart from breaking a lot of bonds > (apparent when looking the model in coot). > I would greatly appreciate some ideas for improving my model > Thanks a lot > D.L -- "I'd jump in myself, if I weren't so good at whistling." Julian, King of Lemurs
Re: [ccp4bb] refinement problem
Dear D.L., I suspect the issue is in determining the correct space group. Sharing the Rmerge and current space group would be useful. Take Care, Sean http://store.p212121.com/
[ccp4bb] refinement problem
Dear all I've got a 3A dataset processed with mosflm and scaled. I've runned MOLREP that gave a solution with a score of 0.52. At this stage the Rfactor was 44%. After my first round of refinement with REFMAC i've got both Rfactor and an Rfree at about 43%. After model building and several refinement rounds i've got an Rfactor 36% and the RFree is 45%. Since there are two molecules in the AU i tried to do NCS phased refinement using NCSREF, but it didn't do anything apart from breaking a lot of bonds (apparent when looking the model in coot). I would greatly appreciate some ideas for improving my model Thanks a lot D.L
Re: [ccp4bb] Refinement problem
You don't mention your space group, it could be one with two distinct "settings." I.e. that a reindexing would be necessary to switch from one set of data to the other. If you scale the two datasets together, is the Rmerge reasonable, or is it over 30%? - === You can't possibly be a scientist if you mind people thinking that you're a fool. - Wonko the Sane === David J. Schuller modern man in a post-modern world MacCHESS, Cornell University schul...@cornell.edu On Sun, 2009-08-09 at 10:39 -0400, kumar wrote: > Dear CCP4ers, > I refined a 3.4 A SAD data in refmac ( using the 'SAD data directly' option > ) and reached a plateau with an Rfree of 36%. Then I tried to extend phases > with the 2.95 A dataset which has no anomalous data (I maintain the same > Rfree flag). The maps look like a moderately good fit, but the Rfree doesn't > go below 39%. > With Autobuild in phenix, map looks good and Rfree went down to 36% but > doesn't change further. Deletion of a 10 residue loop for which no density > is visible increases Rfree. > Model bias is still persistent as sidechains deleted earlier don't come > back. Finally, even with the good density-model fit, there are huge areas of > negative difference density and few areas of positive density. > I also tried a round of simulated annealing in CNS, but it didn't change the > Rfree. Playing with the weights hasn't made much difference either. TLS > refinement also doesn't make any difference. Any ideas how to go about? > Thank you. > Kumar
[ccp4bb] Refinement problem
Dear CCP4ers, I refined a 3.4 A SAD data in refmac ( using the 'SAD data directly' option ) and reached a plateau with an Rfree of 36%. Then I tried to extend phases with the 2.95 A dataset which has no anomalous data (I maintain the same Rfree flag). The maps look like a moderately good fit, but the Rfree doesn't go below 39%. With Autobuild in phenix, map looks good and Rfree went down to 36% but doesn't change further. Deletion of a 10 residue loop for which no density is visible increases Rfree. Model bias is still persistent as sidechains deleted earlier don't come back. Finally, even with the good density-model fit, there are huge areas of negative difference density and few areas of positive density. I also tried a round of simulated annealing in CNS, but it didn't change the Rfree. Playing with the weights hasn't made much difference either. TLS refinement also doesn't make any difference. Any ideas how to go about? Thank you. Kumar
Re: [ccp4bb] Refinement problem
Are you sure of your spacegroup? eleanor Check for Se by doing an anomalous difference map using the PHIC FOM output by refmac. You will have to CAD together the two files - one with the Dano from the data processing with the REFMAC output before doing the map. eleanor Sampath Natarajan wrote: Dear all, Now I'm solving a structure with 1.6A resolution. The data seems good with R-sym (12.4) and all other parameters. Actually the data was collected with SAD phasing. When we checked the data we couldn't find the Se atom in the structure. Since the data resolution is good, we tried to do molecular replacement using Balbes program. It was selected a model with 25% sequence identity and we got the good solution too. I could find all residues in the density and also checked the Ramachandran map which shows almost all residues are in the allowed region. The problem is, I have done refinement many times, the R-factor (45.3) and R-free (51.4) is not reducing during the refinement and also figure of merit is not increasing. Still it remains what I got during the first refinement. The density is also not improving much. Also I could find many cuts in the density. My question is…….. 1.Can we use SAD phasing data for MR solution? 2.Is there any other way to reduce the R/R-free? 3.Why the figure of merit is not increasing even after modeled the residues exactly into the electron density? Thanks, Regards, Sampath
Re: [ccp4bb] Refinement problem
Hi Sampath, On Sat, Jul 26, 2008 at 02:05:17AM +0900, Sampath Natarajan wrote: > Also I could find many cuts in the density. This looks to me like a problem with your low-resolution data? Since you collected 1.6A data my guess is that you probably had a fair amount of overloads. Did you do a low-resolution pass? What is the completeness in the 20-8A range? If there are problems with the low resolution (e.g. lots of very strong reflections missing due to overloads or a large and not well masked beamstop) then it can easily happen that you can't find the heavy atom sites and/or the molecular replacement solution isn't very good. The few low resolution reflections are much more important for the intial structure solution (no matter if experimental phasing or MR) than those thousands of high-resolution reflections ... If you see nice density for a helix but with lot's of gaps (that don't make sense) this could be one of the reasons. 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] Refinement problem
Hi Sampath, Just a couple of other hand-on suggestions about few programs. If you have good completeness upto 1.6A with decent redundancy (>5), experimental phasing is almost always a good idea, depends of course on ordered Se, but hardly ever all of them are disordered, maybe some... how big is the protein? NRes? Matthew Coefficient? Number of Se? a) Which programs did you use for indexing and scaling? If Denzo/Scalepack, you may consider "no merge original index" and "scale anomalous". Also there is a scenario in Denzo manual which can tell you if you have anomalous signal in case you don't have a scan. Also, pay attention to the space group, hklview and pointless (ccp4 prerelease / Phil Evans) can be quite useful for this, both read mtz files. But NCS with screw axis might make life more difficult. b) Use phenix.xtriage to analyse the data, it indicates if there is a possibility of twining. If it gives twin operators, detwin (in ccp4) can be useful even before you do anything with it. c) Try a few different programs, Solve-Resolve might be the quickest, but SHELXD for finding sites, followed by either AutoSHARP or Crank 1.2 (which uses BP3) and PHASER are good candidates. You may need to compare which works better for a given problem..! I have seen examples where odd di-sulphides create lot of problem in finding Se sites.. and PHASER did very well. d) If the raw phases look good, you may try to give it (F, SigF, FOMM, PHI, FreeR and HLs) directly to Arp/Warp. At 1.6, it might work rather well. For refinement, MLHL or SADL might be very useful. Again, at your kind of resolution, Refmac usually does very well.. HTH, Partha On Fri, Jul 25, 2008 at 6:05 PM, Sampath Natarajan <[EMAIL PROTECTED]> wrote: > Dear all, > > > > Now I'm solving a structure with 1.6A resolution. The data seems good with > R-sym (12.4) and all other parameters. Actually the data was collected with > SAD phasing. When we checked the data we couldn't find the Se atom in the > structure. Since the data resolution is good, we tried to do molecular > replacement using Balbes program. It was selected a model with 25% sequence > identity and we got the good solution too. I could find all residues in the > density and also checked the Ramachandran map which shows almost all > residues are in the allowed region. > > > > The problem is, I have done refinement many times, the R-factor (45.3) and > R-free (51.4) is not reducing during the refinement and also figure of merit > is not increasing. Still it remains what I got during the first refinement. > The density is also not improving much. Also I could find many cuts in the > density. > > > > My question is…….. > > > > 1.Can we use SAD phasing data for MR solution? > > 2.Is there any other way to reduce the R/R-free? > > 3.Why the figure of merit is not increasing even after modeled the > residues exactly into the electron density? > > > > Thanks, > > > > Regards, > > > > Sampath > > -- MRC National Institute for Medical Research Division of Molecular Structure The Ridgeway, NW7 1AA, UK Email: [EMAIL PROTECTED] Phone: + 44 208 816 2515
Re: [ccp4bb] Refinement problem
From: Ta Hai Sent: Sat 7/26/2008 3:00 AM To: Sampath Natarajan Subject: RE: [ccp4bb] Refinement problem With the R and freeR factor 45.3% and 51.4%, I'm quite sure that your MR solution is not correct. Although your current model seems fitting with the map, but it's just the biased map after refinement with REFMAC. While now you can't be sure that whether your crystal contain Se atom or not, try MR again: - Make sure you estimate correctly the numbers of molecules/AU. - Prepare the best search model as much as you can - Try with different programs like Molrep, EPMR, Phaser... Good luck, Hai From: CCP4 bulletin board on behalf of Sampath Natarajan Sent: Sat 7/26/2008 2:05 AM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Refinement problem Dear all, Now I'm solving a structure with 1.6A resolution. The data seems good with R-sym (12.4) and all other parameters. Actually the data was collected with SAD phasing. When we checked the data we couldn't find the Se atom in the structure. Since the data resolution is good, we tried to do molecular replacement using Balbes program. It was selected a model with 25% sequence identity and we got the good solution too. I could find all residues in the density and also checked the Ramachandran map which shows almost all residues are in the allowed region. The problem is, I have done refinement many times, the R-factor (45.3) and R-free (51.4) is not reducing during the refinement and also figure of merit is not increasing. Still it remains what I got during the first refinement. The density is also not improving much. Also I could find many cuts in the density. My question is 1.Can we use SAD phasing data for MR solution? 2.Is there any other way to reduce the R/R-free? 3.Why the figure of merit is not increasing even after modeled the residues exactly into the electron density? Thanks, Regards, Sampath
Re: [ccp4bb] Refinement problem
You may need to consider (assuming that space group is correct) 1) Twinning. In this case even wrong solution at the early stages would give lower Rfactor and at the later stages higher Rfactor without accounting for twinning 2) Pseudo translation. In this case the solution could be in wrong origin (you can try zanuda from the YSBLprograms server) 3) you can try arp/warp and/or buccanneer and try to do automatic model building 4) You can try SAD refinement 5) you can check anomolous difference map to see if Se-s are visible 6) If you have NCS you can do averaging Experimental phasing is always a good idea Garib On 25 Jul 2008, at 18:05, Sampath Natarajan wrote: Dear all, Now I'm solving a structure with 1.6A resolution. The data seems good with R-sym (12.4) and all other parameters. Actually the data was collected with SAD phasing. When we checked the data we couldn't find the Se atom in the structure. Since the data resolution is good, we tried to do molecular replacement using Balbes program. It was selected a model with 25% sequence identity and we got the good solution too. I could find all residues in the density and also checked the Ramachandran map which shows almost all residues are in the allowed region. The problem is, I have done refinement many times, the R-factor (45.3) and R-free (51.4) is not reducing during the refinement and also figure of merit is not increasing. Still it remains what I got during the first refinement. The density is also not improving much. Also I could find many cuts in the density. My question is…….. 1.Can we use SAD phasing data for MR solution? 2.Is there any other way to reduce the R/R-free? 3.Why the figure of merit is not increasing even after modeled the residues exactly into the electron density? Thanks, Regards, Sampath
Re: [ccp4bb] Refinement problem
Hi Sampath, You are asking many questions at once. Since I am right now trying to solve a very difficult Se-Met structure, here are some ideas: - Do you have an energy scan on your crystal, showing that there is absorbance at the correct wavelength for Se? If yes, you have proof that there was indeed Se in your crystal; - You describe that you cannot find the Se atom. In theory it is possible that the atom is in the crystal, but not in an ordered fashion and therefore you would not be able to find it. That is theory, I think that in the majority of cases it works fine. You have not told us how the data were collected (synchrotron? wavelength? Inverse beam protocol to optimize SAD?) and whether or not a statistical analysis (with scaleit) tells you if there is a signal there. If you see a signal, then you know Se is present AND ordered. - If you have a MR solution, you can try to use those phases to find the Se atom in a phased anomalous difference map and then (provided that everything is consistent) use a combination of experimental and MR phases; but... - You also did not tell us how you know that your MR solution is correct. Specifically, can you see any features in the structure that are not part of the search model and are sensible? If yes, your solution is useful. If no, you should try omit maps to convince yourself that the MR solution isn't (too) biased and in fact correct - Your statistics given suggest that the MR solution is very poor, with a cons iderable chance that it is not valid. Remember that ~55% R-factor is equivalent to a random solution. If you do refinement and there is no improvement and the statistics are poor, chances are good that the solution wasn't correct in the first place. I worry that you have concluded that since the model fits the density nicely, based on MR, then things must be good, but if you look at Kevin Cowtan's web page with the duck and cat, you will be reminded of the fact that with phases from MR you (almost?) always get nice density from MR, but that does not mean at all that it is correct density. So yes, you should pursue the SAD phases. Remember that the signal will be weak (we cannot judge how weak, depends on the number of Se and size of your problem and the wavelength used), but (good news) you should not have problems with non-isomorphism (since you are not comparing two data sets). The SAD phases, after appropriate density modification, may show you a partial structure - I just tried this for my problem and it did not work for me, but then again, you must try to see if it can be done. Also, assuming that you have native, S-containing protein (as opposed to Se) there is the option of comparing the S- versus Se-protein and you might be able to get phases from that. Finally, others on this forum are better in this matter: 1.6A is fairly high resolution and I wonder if it is possible to pursue direct methods. Probably too low resolution, but I wouldn't know all that w ell, my maps are 5A resolution, so I don't worry about that option. Mark -Original Message- From: Sampath Natarajan <[EMAIL PROTECTED]> To: CCP4BB@JISCMAIL.AC.UK Sent: Fri, 25 Jul 2008 11:05 am Subject: [ccp4bb] Refinement problem Dear all, Now I'm solving a structure with 1.6A resolution. The data seems good with R-sym (12.4) and all other parameters. Actually the data was collected with SAD phasing. When we checked the data we couldn't find the Se atom in the structure. Since the data resolution is good, we tried to do molecular replacement using Balbes program. It was selected a model with 25% sequence identity and we got the good solution too. I could find all residues in the density and also checked the Ramachandran map which shows almost all residues are in the allowed region. The problem is, I have done refinement many times, the R-factor (45.3) and R-free (51.4) is not reducing during the refinement and also figure of merit is not increasing. Still it remains what I got during the first refinement. The density is also not improving much. Also I could find many cuts in the density. My question is…….. 1. Can we use SAD phasing data for MR solution? 2. Is there any other way to reduce the R/R-free? 3. Why the figure of merit is not increasing even after modeled the residues exactly into the electron density? Thanks, Regards, Sampath
[ccp4bb] Refinement problem
Dear all, Now I'm solving a structure with 1.6A resolution. The data seems good with R-sym (12.4) and all other parameters. Actually the data was collected with SAD phasing. When we checked the data we couldn't find the Se atom in the structure. Since the data resolution is good, we tried to do molecular replacement using Balbes program. It was selected a model with 25% sequence identity and we got the good solution too. I could find all residues in the density and also checked the Ramachandran map which shows almost all residues are in the allowed region. The problem is, I have done refinement many times, the R-factor (45.3) and R-free (51.4) is not reducing during the refinement and also figure of merit is not increasing. Still it remains what I got during the first refinement. The density is also not improving much. Also I could find many cuts in the density. My question is…….. 1.Can we use SAD phasing data for MR solution? 2.Is there any other way to reduce the R/R-free? 3.Why the figure of merit is not increasing even after modeled the residues exactly into the electron density? Thanks, Regards, Sampath
Re: [ccp4bb] refinement problem
Hi, It would be nice if you mention more clearly how you have done TLS..i.e.,domains or individual residues or molecules !! You should keep in mind that you have 2.8A data. I would do TLSANL with domains or molecules. Best regards, Krishna Ch PhD Student Hannover Medical school Germany On Wed, May 14, 2008 at 11:36 AM, parkash <[EMAIL PROTECTED]> wrote: > Hi, > I have one structural refinement problem. > I am working on a protein crystals which diffracted to 2.8 Å. But when I > refine through REFMAC5, with 0.1 wt(geometry to x-ray terms), I get high > B-factors around 70. But if I do TLS refinement, the R-factors lower down > and B-factors come down to 40. > > My question is; Are TLS refined B-factors real?? Can I trust them? > or what else could be done to lower down the B.factors? > > BR > vimal >
Re: [ccp4bb] refinement problem
parkash wrote: Hi, I have one structural refinement problem. I am working on a protein crystals which diffracted to 2.8 Å. But when I refine through REFMAC5, with 0.1 wt(geometry to x-ray terms), I get high B-factors around 70. But if I do TLS refinement, the R-factors lower down and B-factors come down to 40. My question is; Are TLS refined B-factors real?? Can I trust them? or what else could be done to lower down the B.factors? BR vimal This is or should be a FAQ - all B factors after TLS are relative to the TLS restraints Try running TLSANL and that should give you back a more meaningful overall B Eleanor
[ccp4bb] refinement problem
Hi, I have one structural refinement problem. I am working on a protein crystals which diffracted to 2.8 Å. But when I refine through REFMAC5, with 0.1 wt(geometry to x-ray terms), I get high B-factors around 70. But if I do TLS refinement, the R-factors lower down and B-factors come down to 40. My question is; Are TLS refined B-factors real?? Can I trust them? or what else could be done to lower down the B.factors? BR vimal
Re: [ccp4bb] refinement problem of labelled RNA complex
Hi, It's much easier to begin with the cif file created by Refmac5 itself, but this cif file needs to be checked and corrected manually for the restrains. Certainly, Iodine is not happy with the current restrains. Most of the time, i have found problems beginning with the protonation and the bond order and then extending to the geometrical freedom of the molecule. You may want to try Monomer sketcher, where you can manipulate the monomers visually. all the best Manish - Center for Advanced Research in Biotechnology University of Maryland Biotechnology Institute 9600 Gudelsky Drive, Rockville MD 20850 USA Tel: +1-240-314-6130; fax: +1-240-314-6255 --- ºîº£·å <[EMAIL PROTECTED]> wrote: > Dear all; > I am a fresher of Refmac5. And now I'm trying to refine a protein and > labelled RNA complex using > Refmac. It's a Iodo-U labelled RNA. Unfortunately, these is no appropriate > lib file in the > current refmac dictionaries. So I use cif file which created by Refmc itself, > surprisingly, the > base of uridine distorted greatly. Could someone like to tell me how to make > the appropriate cif > file to solve this problem. Another question is that the RNA has a phosphate > group in 5', How > to define this in the cif file? > > Could someone help me? Thanks a lot! >Best wishes, > >Hai-Feng Hou > > Biological macromolecule group > Beijing Synchrotron Radiation Facility > Institute of High Energy Physics > Chinese Academy of Sciences > Beijing, China > Tel: +86-10-88234208 > E-mail: [EMAIL PROTECTED] > >2008-02-18 > Now you can chat without downloading messenger. Go to http://in.messenger.yahoo.com/webmessengerpromo.php
[ccp4bb] refinement problem of labelled RNA complex
Dear all; I am a fresher of Refmac5. And now I'm trying to refine a protein and labelled RNA complex using Refmac. It's a Iodo-U labelled RNA. Unfortunately, these is no appropriate lib file in the current refmac dictionaries. So I use cif file which created by Refmc itself, surprisingly, the base of uridine distorted greatly. Could someone like to tell me how to make the appropriate cif file to solve this problem. Another question is that the RNA has a phosphate group in 5', How to define this in the cif file? Could someone help me? Thanks a lot! Best wishes, Hai-Feng Hou Biological macromolecule group Beijing Synchrotron Radiation Facility Institute of High Energy Physics Chinese Academy of Sciences Beijing, China Tel: +86-10-88234208 E-mail: [EMAIL PROTECTED] 2008-02-18
[ccp4bb] refinement problem of labelled RNA complex
Dear all; I am a fresher of Refmac5. And now I'm trying to refine a protein and labelled RNA complex using Refmac. It's a Iodo-U labelled RNA. Unfortunately, these is no appropriate lib file in the current refmac dictionaries. So I use cif file which created by Refmc itself, surprisingly, the base of uridine distorted greatly. Could someone like to tell me how to make the appropriate cif file to solve this problem. Another question is that the RNA has a phosphate group in 5', How to define this in the cif file? Could someone help me? Thanks a lot! Best wishes, Hai-Feng Hou Biological macromolecule group Beijing Synchrotron Radiation Facility Institute of High Energy Physics Chinese Academy of Sciences Beijing, China Tel: +86-10-88234208 E-mail: [EMAIL PROTECTED] 2008-02-18
