Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
There were many complaints that Phaser was too strict in its packing check, so since version 2.3 (the one with the soon-to-be-updated version of CCP4), Phaser has by default allowed up to 5% of the trace atoms to clash. In more recent versions (available with Phenix and soon from CCP4), if something is rejected for packing but it otherwise looks like a good potential solution (TFZ8), then a warning message is triggered and summarised at the end of the logfile. In our tests the 5% cutoff seems to be a good default, but in cases with low sequence identity (where there are likely to be loops that are very different) we almost always use a tool like our sculptor program or chainsaw in CCP4 to trim off bits that don't align with the target sequence. So I would agree with the suggestion to trim the model. However, in our hands a polyAla model is almost always worse than one where the identical side chains are left intact, and other residues are trimmed at most to the gamma atom. We're always interested in feedback, so if there are still cases where Phaser rejects good solutions we'd like to hear about them! Regards, Randy Read On 1 May 2012, at 00:18, Ho Leung Ng wrote: When searching for multiple molecules/ASU, you need to be careful with how the software handles packing. Small but acceptable clashes can accumulate and cause the searches to fail. I suggest using a highly trimmed as well as a poly-alanine model. I've had success with both epmr and Phaser. With Phaser, you'll probably want to loosen the acceptable number of packing clashes. Ho Ho Leung Ng University of Hawaii at Manoa Assistant Professor, Department of Chemistry h...@hawaii.edu On Mon, Apr 30, 2012 at 1:00 PM, CCP4BB automatic digest system lists...@jiscmail.ac.uk wrote: Date:Mon, 30 Apr 2012 15:41:54 + From:Ke, Jiyuan jiyuan...@vai.org Subject: Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503 -- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills RoadE-mail: rj...@cam.ac.uk Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
Well - I have found lots of molecules but usually not in a single run. The first thing to think about is: is this likely to be a dimer? trimer? tetramer? Things to consider - a) any non-cryst translation? b) tthe self rotation might give a clue - c) is the model a multimer, c)what do the biochemists suggest? etc etc - If you are looking for a dimer say - the try searching with it.. Eleanor On 30 April 2012 17:33, David Schuller dj...@cornell.edu wrote: On 04/30/12 11:41, Ke, Jiyuan wrote: Dear All, ** ** I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! ** ** I solved 8 copies with 32% identity to the search model (1ZL9). I used BEAST for the MR, in the days before Phaser. You say you have a single protein, which I interpret to mean that the active unit is a monomer. Bummer, searching with a multimer would certainly simplify things. See that your search model is compact. Is the search model a single domain? Does the sequence similarity extend throughout the entire thing? Are there loops which should be clipped? With such a large number of monomers, translation-only NCS seems likely, so be on the lookout for that. -- === All Things Serve the Beam === David J. Schuller modern man in a post-modern world MacCHESS, Cornell University schul...@cornell.edu
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
For those of you who want to attempt multi-copy problems with Phaser... Most of the work on Phaser over the last year has gone into improving results for multi-copy problems, to increase the signal-to-noise of the searches and reduce the computation time (more solutions for less cpu). The biggest change is the addition of a likelihood treatment of translational NCS, which is common for multi-copy problems. However, there are also changes to the search algorithm and packing function that make a large impact. For example, the packing function was performing poorly when there were 20 copies in the asymmetric unit because if 19 were placed and not overlapped then the placement of the 20th molecule could result in total overlap, because 5% of the total number of residues was allowed in clashes but 5% of the residues was one whole molecule. So there was, in effect, no packing test just when the packing test should have been providing good discrimination. This has been fixed, Phaser-2.5.0 (nightly builds) has these additional features to help in multi-copy cases: Packing function changes as mentioned above; Translational NCS correction; TNCS/twin detection; Search ensembles that are multimers with pointgroup symmetry can be placed on crystallographic symmetry axes without triggering packing clashes; Search algorithm can amalgamate multiple copies found in one translation function (prevents unneccessary branching); Space group not determined by placement of first ensemble, but carried through until solution is found; Refinement of overall B-factor for each placed copy (can vary widely for different copies in the ASU); Refinement of ensemble variance when signal-to-noise is low (true for placement of first few copies in multi-copy problems), Phaser Tip #1: Search for all the copies you think are present in the ASU in one go. Phaser Tip #2: If you find e.g. 15 in one run, you can re-start Phaser from this point using the .sol file to look for more. Phaser Tip #3: Remember to use the appropriate composition for the number you think are present in any given run (not necessarily the same as the number you are searching for). Since this code is in active development, at the moment we are particularly interested in multi-copy cases that don't solve. Airlie On 04/30/12 11:41, Ke, Jiyuan wrote: Dear All, ** ** I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! ** **
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
This seems to be very a common scenario in MR and recently we have added some code to phaser.MRage that would handle it automatically. Each partial solution is analysed for the presence of (complete of incomplete) assemblies that obey local point group symmetry. After such an assembly is identified, the program can either fill in molecules where they should be (provided the obey the previously identified local symmetry), and can also search with the full assembly. This greatly enhances the signal (since the model comprises a larger fraction of the asymmetric unit), and also speeds up the search. (MRage is not yet included in the current CCP4 release, but available as alpha-test in recent Phenix nightly builds.) BW, Gabor On Apr 30 2012, Roger Rowlett wrote: A partial solution can potentially lead you to an appropriate MR solution. With many protein chains in the ASU, there will be several reasonable possibilities by Matthews analysis. When I originally solved 2A8D, cell content analysis suggested 8 monomers per ASU, but it was clear after a few MR runs that was not going to work. Inspecting the packing of a partial solution with 4 monomers, which formed a nice, biological-looking tetramer, clearly suggested that another dimer would fit into the lattice just nicely. A search with 3 such dimers produced an excellent MR solution starting point for the final refinement. I used a similar procedure for 3UAO. The Matthews analysis suggested 10-12 chains per ASU, but it was clearly 8 based on packing of partial solutions. A 4-dimer search was immediately successful. Cheers, ___ Roger S. Rowlett Gordon Dorothy Kline Professor Department of Chemistry Colgate University 13 Oak Drive Hamilton, NY 13346 tel: (315)-228-7245 ofc: (315)-228-7395 fax: (315)-228-7935 email: rrowl...@colgate.edu On 4/30/2012 5:20 PM, mjvdwo...@netscape.net wrote: Provided that you guess the number of copies and your guess is reasonably close, my experience is that Phaser will do the job. But you have to tell it how many copies you expect, or it will never make sense of the data. When I did my structure with 6(?) copies some years ago, I guessed a number that was close enough and then when I inspected the electron density I could see that there were more copies than I had told the software and all was fine after that. It was surprising to see that good solutions were obvious from a packing consideration, while inadequate solutions were obviously wrong. Mark -Original Message- From: Ke, Jiyuan jiyuan...@vai.org To: CCP4BB CCP4BB@JISCMAIL.AC.UK Sent: Mon, Apr 30, 2012 2:28 pm Subject: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503 -- ## Dr Gabor Bunkoczi Cambridge Institute for Medical Research Wellcome Trust/MRC Building Addenbrooke's Hospital Hills Road Cambridge CB2 0XY ##
[ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
For large copy number MR solutions, we have found that EPMR is a good alternative to Phaser when the latter doesn't find a solution. However, we also have noted in some experimental testing that both Phaser and EPMR have some difficulties with copy numbers greater than 4-6. Another alternative, if you have lots of patience, is Queen of Spades. It can sometimes place large numbers of models in the ASU at the cost of a loong wait. But a CS collaborator and I found that genetic algorithms (e.g. EPMR) were pretty efficient at large copy number searches. If you can place a handful of protein chains in the ASU, you might be able to get away with searching with some dimer or tetramer units from that partial solution. That's what I wound up doing with a 6-chains-per-ASU problem a few years ago (PDB 2A8D). The 3-dimer search in EPMR worked like a charm, whereas the 6-monomer search was hopeless at the time. Cheers, ___ Roger S. Rowlett Gordon Dorothy Kline Professor Department of Chemistry Colgate University 13 Oak Drive Hamilton, NY 13346 tel: (315)-228-7245 ofc: (315)-228-7395 fax: (315)-228-7935 email: rrowl...@colgate.edu On 4/30/2012 11:41 AM, Ke, Jiyuan wrote: Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
On 04/30/12 11:41, Ke, Jiyuan wrote: Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! I solved 8 copies with 32% identity to the search model (1ZL9). I used BEAST for the MR, in the days before Phaser. You say you have a single protein, which I interpret to mean that the active unit is a monomer. Bummer, searching with a multimer would certainly simplify things. See that your search model is compact. Is the search model a single domain? Does the sequence similarity extend throughout the entire thing? Are there loops which should be clipped? With such a large number of monomers, translation-only NCS seems likely, so be on the lookout for that. -- === All Things Serve the Beam === David J. Schuller modern man in a post-modern world MacCHESS, Cornell University schul...@cornell.edu
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
Dear Jiyuan have you run any data diagnostics on your dataset? Ccp4 offers quite some options (truncate, detwin etc), and xtriage from PHENIX is also very powerful. And how sure are you of your laue group symmetry and space group (pointless via ccp4 and xtriage via PHENIX can be very helpful)? Also, is your data strongly anisotropic? Furthermore, the presence of pseudotranslation may not be an unlikely scenario given your large unit cell and large number of copies in the asu. MolRep in ccp4 can deal well with that, and is worth trying anyway regardless of pseudotranslation issues. Also I could not help notice that your c axis is b+b/2. best regards Savvas On 30 Apr 2012, at 17:41, Ke, Jiyuan wrote: Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
Provided that you guess the number of copies and your guess is reasonably close, my experience is that Phaser will do the job. But you have to tell it how many copies you expect, or it will never make sense of the data. When I did my structure with 6(?) copies some years ago, I guessed a number that was close enough and then when I inspected the electron density I could see that there were more copies than I had told the software and all was fine after that. It was surprising to see that good solutions were obvious from a packing consideration, while inadequate solutions were obviously wrong. Mark -Original Message- From: Ke, Jiyuan jiyuan...@vai.org To: CCP4BB CCP4BB@JISCMAIL.AC.UK Sent: Mon, Apr 30, 2012 2:28 pm Subject: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
Why not run a number of jobs in parallel with varying numbers of monomers/asu? JPK On Mon, Apr 30, 2012 at 4:20 PM, mjvdwo...@netscape.net wrote: Provided that you guess the number of copies and your guess is reasonably close, my experience is that Phaser will do the job. But you have to tell it how many copies you expect, or it will never make sense of the data. When I did my structure with 6(?) copies some years ago, I guessed a number that was close enough and then when I inspected the electron density I could see that there were more copies than I had told the software and all was fine after that. It was surprising to see that good solutions were obvious from a packing consideration, while inadequate solutions were obviously wrong. Mark -Original Message- From: Ke, Jiyuan jiyuan...@vai.org To: CCP4BB CCP4BB@JISCMAIL.AC.UK Sent: Mon, Apr 30, 2012 2:28 pm Subject: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503 -- *** Jacob Pearson Keller Northwestern University Medical Scientist Training Program email: j-kell...@northwestern.edu ***
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
A partial solution can potentially lead you to an appropriate MR solution. With many protein chains in the ASU, there will be several reasonable possibilities by Matthews analysis. When I originally solved 2A8D, cell content analysis suggested 8 monomers per ASU, but it was clear after a few MR runs that was not going to work. Inspecting the packing of a partial solution with 4 monomers, which formed a nice, biological-looking tetramer, clearly suggested that another dimer would fit into the lattice just nicely. A search with 3 such dimers produced an excellent MR solution starting point for the final refinement. I used a similar procedure for 3UAO. The Matthews analysis suggested 10-12 chains per ASU, but it was clearly 8 based on packing of partial solutions. A 4-dimer search was immediately successful. Cheers, ___ Roger S. Rowlett Gordon Dorothy Kline Professor Department of Chemistry Colgate University 13 Oak Drive Hamilton, NY 13346 tel: (315)-228-7245 ofc: (315)-228-7395 fax: (315)-228-7935 email: rrowl...@colgate.edu On 4/30/2012 5:20 PM, mjvdwo...@netscape.net wrote: Provided that you guess the number of copies and your guess is reasonably close, my experience is that Phaser will do the job. But you have to tell it how many copies you expect, or it will never make sense of the data. When I did my structure with 6(?) copies some years ago, I guessed a number that was close enough and then when I inspected the electron density I could see that there were more copies than I had told the software and all was fine after that. It was surprising to see that good solutions were obvious from a packing consideration, while inadequate solutions were obviously wrong. Mark -Original Message- From: Ke, Jiyuan jiyuan...@vai.org To: CCP4BB CCP4BB@JISCMAIL.AC.UK Sent: Mon, Apr 30, 2012 2:28 pm Subject: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
When searching for multiple molecules/ASU, you need to be careful with how the software handles packing. Small but acceptable clashes can accumulate and cause the searches to fail. I suggest using a highly trimmed as well as a poly-alanine model. I've had success with both epmr and Phaser. With Phaser, you'll probably want to loosen the acceptable number of packing clashes. Ho Ho Leung Ng University of Hawaii at Manoa Assistant Professor, Department of Chemistry h...@hawaii.edu On Mon, Apr 30, 2012 at 1:00 PM, CCP4BB automatic digest system lists...@jiscmail.ac.uk wrote: Date: Mon, 30 Apr 2012 15:41:54 + From: Ke, Jiyuan jiyuan...@vai.org Subject: Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
Re: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit
We solved recently a structure with 18 dimers in the a.u. related by translational symmetry. Molrep did a great job at quickly placing all molecules but few. The missing ones were found later by map inspection. We found rather important to play with the resolution cutoff. I understand new versions of Phaser can also handle pseudotranslation very effectively. Good luck. R From: CCP4 bulletin board [CCP4BB@JISCMAIL.AC.UK] On Behalf Of Ke, Jiyuan [jiyuan...@vai.org] Sent: Monday, April 30, 2012 4:41 PM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Suggestions for solving a structure with 8-10 copies per asymmetric unit Dear All, I have a question regarding solving a crystal structure by molecular replacement. It is a single protein with a molecular weight of 25.5 kDa. The cell dimension is rather big from the diffraction data ( 90.9 Å, 143.9 Å, 216.3Å, 90°, 90°, 90°). The possible space group is P212121. With such a big unit cell, we predicted that there are 8-10 molecules per asymmetric unit. We have a decent model with sequence similarity of 49%. I tried several times with Phaser search with the current model and had difficulty to find any clear solution. Has anyone seen such cases and any suggestions to solve the structure? Thanks! Jiyuan Ke, Ph.D. Research Scientist Van Andel Research Institute 333 Bostwick Ave NE Grand Rapids, MI 49503
