[ccp4bb] Structure idealisation Refmac_5.5.0072
Dear, I want to optimize a DNA-helix with the function "Structure idealisation" in Refmac_5.5.00782 (CCP4_6.1.1). My question, is this performing just a geometry optimization (against a library), or is there also an energy-optimization of some kind involved,..? And according to the number of cycles (default 10) used, different structural results are obtained, hence is there a means of estimating the ideal number of cycles to use..? Many thanks Kristof Van Hecke -- Kristof Van Hecke, PhD Biomoleculaire Architectuur Celestijnenlaan 200 F B-3001 Heverlee (Leuven) Tel: +32(0)16327477 -- Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
Re: [ccp4bb] Structure idealisation Refmac_5.5.0072
Hi Kristof The trivial, and it turns out only, answer is "as many as it takes to converge". Unfortunately for everyone carrying out optimisations, or indeed almost any kind of computation, Alan Turing proved (1936) - Google for "Turing halting problem" - that a general algorithm to solve the halting problem for all possible program-input pairs cannot exist, or to put it another way, there's no way to know in advance what a program will do without actually running it. The number of iterations required will obviously depend on how far away the starting parameters are from the optimum which clearly cannot be known in advance. In any case, if you don't like the result that the program produces then the answer doesn't lie in changing the number of iterations: rather the answer lies in changing the input and starting conditions, since an optimisation procedure is valid only if carried through to convergence (otherwise the starting assumptions are not satisfied), i.e. taking as many (or as few) iterations as it needs to satisfy the stopping criteria. Of course you could change the stopping criteria but they determine how accurate the result will be, so the question then would be "how accurate do you want it?". So much for the theory, on a practical note I've noticed that for some reason structure idealisation with Refmac requires a very large number of iterations to converge (I have not tried it with other refinement programs though, maybe I should), given that the stopping criteria are that the RMSDs for bonds and angles should be essentially zero (they can only be non-zero in the case that the restraints are inconsistent with each other, e.g. rings don't close under the restraints), so maybe someone can answer why that is, it has baffled me for a while. Cheers -- Ian > -Original Message- > From: owner-ccp...@jiscmail.ac.uk [mailto:owner-ccp...@jiscmail.ac.uk] On > Behalf Of Kristof Van Hecke > Sent: 25 March 2009 10:42 > To: bulletin_ccp4 > Subject: Structure idealisation Refmac_5.5.0072 > > Dear, > > I want to optimize a DNA-helix with the function "Structure idealisation" > in Refmac_5.5.00782 (CCP4_6.1.1). > My question, is this performing just a geometry optimization (against a > library), or is there also an energy-optimization of some kind > involved,..? > > And according to the number of cycles (default 10) used, different > structural results are obtained, hence is there a means of estimating the > ideal number of cycles to use..? > > Many thanks > > Kristof Van Hecke > > -- > Kristof Van Hecke, PhD > Biomoleculaire Architectuur > Celestijnenlaan 200 F > B-3001 Heverlee (Leuven) > Tel: +32(0)16327477 > -- > > > > > > Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm for more > information. > Disclaimer This communication is confidential and may contain privileged information intended solely for the named addressee(s). It may not be used or disclosed except for the purpose for which it has been sent. If you are not the intended recipient you must not review, use, disclose, copy, distribute or take any action in reliance upon it. If you have received this communication in error, please notify Astex Therapeutics Ltd by emailing i.tic...@astex-therapeutics.com and destroy all copies of the message and any attached documents. Astex Therapeutics Ltd monitors, controls and protects all its messaging traffic in compliance with its corporate email policy. The Company accepts no liability or responsibility for any onward transmission or use of emails and attachments having left the Astex Therapeutics domain. Unless expressly stated, opinions in this message are those of the individual sender and not of Astex Therapeutics Ltd. The recipient should check this email and any attachments for the presence of computer viruses. Astex Therapeutics Ltd accepts no liability for damage caused by any virus transmitted by this email. E-mail is susceptible to data corruption, interception, unauthorized amendment, and tampering, Astex Therapeutics Ltd only send and receive e-mails on the basis that the Company is not liable for any such alteration or any consequences thereof. Astex Therapeutics Ltd., Registered in England at 436 Cambridge Science Park, Cambridge CB4 0QA under number 3751674
Re: [ccp4bb] Structure idealisation Refmac_5.5.0072
Dear Ian, If one has sufficiently high resolution data, full matrix refinement with SHELXL might be the best way to get real convergence. The installation of the program also converges faster (zero dependence!). Best wishes, George Prof. George M. Sheldrick FRS Dept. Structural Chemistry, University of Goettingen, Tammannstr. 4, D37077 Goettingen, Germany Tel. +49-551-39-3021 or -3068 Fax. +49-551-39-22582 On Wed, 25 Mar 2009, Ian Tickle wrote: > Hi Kristof > > The trivial, and it turns out only, answer is "as many as it takes to > converge". Unfortunately for everyone carrying out optimisations, or > indeed almost any kind of computation, Alan Turing proved (1936) - > Google for "Turing halting problem" - that a general algorithm to solve > the halting problem for all possible program-input pairs cannot exist, > or to put it another way, there's no way to know in advance what a > program will do without actually running it. > > The number of iterations required will obviously depend on how far away > the starting parameters are from the optimum which clearly cannot be > known in advance. In any case, if you don't like the result that the > program produces then the answer doesn't lie in changing the number of > iterations: rather the answer lies in changing the input and starting > conditions, since an optimisation procedure is valid only if carried > through to convergence (otherwise the starting assumptions are not > satisfied), i.e. taking as many (or as few) iterations as it needs to > satisfy the stopping criteria. Of course you could change the stopping > criteria but they determine how accurate the result will be, so the > question then would be "how accurate do you want it?". > > So much for the theory, on a practical note I've noticed that for some > reason structure idealisation with Refmac requires a very large number > of iterations to converge (I have not tried it with other refinement > programs though, maybe I should), given that the stopping criteria are > that the RMSDs for bonds and angles should be essentially zero (they can > only be non-zero in the case that the restraints are inconsistent with > each other, e.g. rings don't close under the restraints), so maybe > someone can answer why that is, it has baffled me for a while. > > Cheers > > -- Ian > > > -Original Message- > > From: owner-ccp...@jiscmail.ac.uk [mailto:owner-ccp...@jiscmail.ac.uk] > On > > Behalf Of Kristof Van Hecke > > Sent: 25 March 2009 10:42 > > To: bulletin_ccp4 > > Subject: Structure idealisation Refmac_5.5.0072 > > > > Dear, > > > > I want to optimize a DNA-helix with the function "Structure > idealisation" > > in Refmac_5.5.00782 (CCP4_6.1.1). > > My question, is this performing just a geometry optimization (against > a > > library), or is there also an energy-optimization of some kind > > involved,..? > > > > And according to the number of cycles (default 10) used, different > > structural results are obtained, hence is there a means of estimating > the > > ideal number of cycles to use..? > > > > Many thanks > > > > Kristof Van Hecke > > > > -- > > Kristof Van Hecke, PhD > > Biomoleculaire Architectuur > > Celestijnenlaan 200 F > > B-3001 Heverlee (Leuven) > > Tel: +32(0)16327477 > > -- > > > > > > > > > > > > Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm for more > > information. > > > > > > Disclaimer > This communication is confidential and may contain privileged information > intended solely for the named addressee(s). It may not be used or disclosed > except for the purpose for which it has been sent. If you are not the > intended recipient you must not review, use, disclose, copy, distribute or > take any action in reliance upon it. If you have received this communication > in error, please notify Astex Therapeutics Ltd by emailing > i.tic...@astex-therapeutics.com and destroy all copies of the message and any > attached documents. > Astex Therapeutics Ltd monitors, controls and protects all its messaging > traffic in compliance with its corporate email policy. The Company accepts no > liability or responsibility for any onward transmission or use of emails and > attachments having left the Astex Therapeutics domain. Unless expressly > stated, opinions in this message are those of the individual sender and not > of Astex Therapeutics Ltd. The recipient should check this email and any > attachments for the presence of computer viruses. Astex Therapeutics Ltd > accepts no liability for damage caused by any virus transmitted by this > email. E-mail is susceptible to data corruption, interception, unauthorized > amendment, and tampering, Astex Therapeutics Ltd only send and receive > e-mails on the basis that the Company is not liable for any such alteration > or any consequences thereof. > Astex Therapeutics Ltd., Registered in England at 436 Cambridge Science
Re: [ccp4bb] Structure idealisation Refmac_5.5.0072
Kristof Van Hecke wrote: I want to optimize a DNA-helix with the function "Structure idealisation" in Refmac_5.5.00782 (CCP4_6.1.1). My question, is this performing just a geometry optimization (against a library), or is there also an energy-optimization of some kind involved,..? What's the difference? And according to the number of cycles (default 10) used, different structural results are obtained, hence is there a means of estimating the ideal number of cycles to use..? I generally keep re-inputting the output PDB back into refmac until the input and output PDB files are "identical" to, say, withing 0.001 A. This is my definition of "convergence". It is important to use a small number of cycles for each run because if you don't atoms can move far enough to start "bumping" into atoms that were not close enough to trigger a "bump" restraint at the beginning. At least, this is how it was back in "my day". -James Holton MAD Scientist
Re: [ccp4bb] Structure idealisation Refmac_5.5.0072
But aren't the bump restraints re-evaluated at each iteration? That was my understanding (but I could be wrong), after all it's not a big deal to compute interatomic distances. One contribution that is not re-evaluated every iteration but only at iteration 1 is (I believe, but again I could be wrong) the solvent contribution, I guess because the assumption is that it doesn't change much. All the same it's interesting that you mention that idea, because I have been trying precisely that with regular refinement and in some cases I get significantly lower final R-factors for the same total number of iterations if I split it into a series of short jobs of maybe only 3 or 4 iterations each (doesn't help every time though), each time feeding the output PDB from one job into the next one as you describe. I've no idea why it works though! Cheers -- Ian > -Original Message- > From: owner-ccp...@jiscmail.ac.uk [mailto:owner-ccp...@jiscmail.ac.uk] On > Behalf Of James Holton > Sent: 26 March 2009 01:08 > To: Kristof Van Hecke > Cc: CCP4BB@jiscmail.ac.uk > Subject: Re: [ccp4bb] Structure idealisation Refmac_5.5.0072 > > Kristof Van Hecke wrote: > > I want to optimize a DNA-helix with the function "Structure > > idealisation" in Refmac_5.5.00782 (CCP4_6.1.1). > > My question, is this performing just a geometry optimization (against > > a library), or is there also an energy-optimization of some kind > > involved,..? > What's the difference? > > > > And according to the number of cycles (default 10) used, different > > structural results are obtained, hence is there a means of estimating > > the ideal number of cycles to use..? > > > I generally keep re-inputting the output PDB back into refmac until the > input and output PDB files are "identical" to, say, withing 0.001 A. > This is my definition of "convergence". It is important to use a small > number of cycles for each run because if you don't atoms can move far > enough to start "bumping" into atoms that were not close enough to > trigger a "bump" restraint at the beginning. At least, this is how it > was back in "my day". > > -James Holton > MAD Scientist Disclaimer This communication is confidential and may contain privileged information intended solely for the named addressee(s). It may not be used or disclosed except for the purpose for which it has been sent. If you are not the intended recipient you must not review, use, disclose, copy, distribute or take any action in reliance upon it. If you have received this communication in error, please notify Astex Therapeutics Ltd by emailing i.tic...@astex-therapeutics.com and destroy all copies of the message and any attached documents. Astex Therapeutics Ltd monitors, controls and protects all its messaging traffic in compliance with its corporate email policy. The Company accepts no liability or responsibility for any onward transmission or use of emails and attachments having left the Astex Therapeutics domain. Unless expressly stated, opinions in this message are those of the individual sender and not of Astex Therapeutics Ltd. The recipient should check this email and any attachments for the presence of computer viruses. Astex Therapeutics Ltd accepts no liability for damage caused by any virus transmitted by this email. E-mail is susceptible to data corruption, interception, unauthorized amendment, and tampering, Astex Therapeutics Ltd only send and receive e-mails on the basis that the Company is not liable for any such alteration or any consequences thereof. Astex Therapeutics Ltd., Registered in England at 436 Cambridge Science Park, Cambridge CB4 0QA under number 3751674