I go along with all Eckhard's cautionary advice & particularly his suggestion that all restraint info be deposited, as without that it's impossible to reproduce the experiment! - the restraint info (both target values & weights or s.d.'s) is literally just as important as the X-ray data. The only problem is that it's not possible to do that under the present scheme of depositing co-ordinates & structure factors in the PDB (or is it?).
Additionally I would point out that it's impossible to say, as Huiying appears to be doing, whether differences in Rfree (or worse still, Rwork) between models are significant: there is simply no valid test of significance for the R factor statistic in this situation. For that in my view you need to look at the free likelihood (or rather its log) which the latest versions of Refmac now print out in the summary table at the end of each successful run. IMO the limiting values of the 'Bayes factor' as suggested by Jeffreys (see for example http://en.wikipedia.org/wiki/Bayes_factor) are suitable for this purpose. I've adapted Jeffreys' table to express the 'weight of evidence' for one model (M1) relative to the alternative (M2) as a difference between the log(free likelihood) in units of 'nats', i.e. units of the natural log (base e) of the Bayes factor K (which is essentially a likelihood ratio): K nats Weight of evidence = ==== ================== < 1:1 < 0 Negative (supports M2) 1:1 to 3:1 0 to 1.1 Barely worth mentioning 3:1 to 10:1 1.1 to 2.3 Substantial 10:1 to 30:1 2.3 to 3.4 Strong 30:1 to 100:1 3.4 to 4.6 Very strong > 100:1 > 4.6 Decisive Note that you need to look at the log of the *free* likelihood (corresponding to Rfree) *not* the working set log likelihood (corresponding to Rwork, and which also appears in the Refmac summary table). Also it's important to note that all this assumes that the refinement against the working set has converged (i.e. no significant 'contamination' of the model by the test set), otherwise the test set statistics are invalid. This is of course not intended in any way to be a substitute for model-based validation (Ramachandran plots etc), but if anyone has a better idea for quantifying the degree to which the X-ray data supports one model over an alternative one I'd like to hear it! Cheers -- Ian > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Eckhard Hofmann > Sent: 06 December 2008 09:26 > To: CCP4BB@JISCMAIL.AC.UK > Cc: [EMAIL PROTECTED] > Subject: Re: Summary - torsion angle restraints in REFMAC > > Hi Huiying, > I would be very careful to play around with the overall > weights for your > purpose. Keep in mind that the overall weights will affect the full > model. Usually the weights on proteins will be well balanced > between the > different types of restraints, and you will be meddling with this > balance. While it will probably not be "visible" in the structure, it > nevertheless is not the appropriate way to do this and will lead to > small, but systematical deviations. You definitely should be > using the > sigmas in the cif to adjust the weight for the individual torsion you > are worried about. > > I expect you always do a dry run with the isolated ligand in refmac > "structure idealization" mode after fiddling with the cif, as this is > most important to make sure you don't do something wrong > there (maybe a > typo or similar)? Especially with ligand cifs the golden rule > "garbage > in, garbage out" applies, as can be found in many pdb structures. > > This would be the point to urge all depositors to also put the > additional restraint files into the database as auxiliary > files to their > structures. > > Cheers, > Eckhard > > > > Huiying Li schrieb: > > > > > > The keyword script Eckhard mentioned did the job I wanted > to achieve. I > > created the cif file for my ligand through the Monomer > Library Stetcher > > and then edited torsion angle part to include the restraint > for the OH > > tiling angle. I am running the CCP4 Suite 6.0.2 and CCP4i > 1.4.4.2. but I > > still have to include an external keyword script > > restr tors include resi XXX > > to impose torsion angle restraint on the ligand. With the > script read in > > I can further adjust the Overall wt for Torsion under Geometric > > parameters menu to achieve a geometry that best fits to the density. > > > > I have tried weighting term 0.05 or 0.08 for this case. The smaller > > weight at 0.05 gave me better planarity for the aromatic residues > > without sacrifying too much in R/R-free (only 0.1% lower > with 0.08 weight). > > > > Thanks again for suggestions from many others. > > > > Huiying > > > > On Fri, 5 Dec 2008, Eckhard Hofmann wrote: > > > >> Hi Huiying, > >> could you tell us which version of refmac you are using? > >> There has been an issue with torsion definitions in ligands NEVER > >> included in the refinement, even if they were in the cif > (at least up > >> to Refmac_5.2.0019). I did have the same effect you had > and wondered > >> why. (See list in October 2007) > >> > >> Garib modified the code in a later version of 5.4 to allow you to > >> include all torsions from one residues by applying the > following line: > >> > >> restr tors include resi XXX > >> or to include only one specific definition in refinement > >> restr tors include resi XXX name var_1 > >> > >> In CCP4i you'll have to include this with the Developers Option to > >> include an external keyword-file. NOTE: Without doing > anything, the > >> torsion definitions are still NOT used, which would explain the > >> behaviour you observe. This is a feature, not a bug .... > >> > >> I'am not sure what the situation is in the current version, maybe > >> Garib should answer this? > >> > >> Hope that helps, > >> Eckhard > >> > >> > >> > >> Huiying Li schrieb: > >>> I want to impose restraints during REFMAC refinement on > the tortion > >>> angles that control the tilting of an OH group from a plane in a > >>> ligand bound to the protein. A few things that confused me: > >>> > >>> 1. In library cif file, should I just increase or decrease the > >>> tor.value_angle_esd if I want to loosen or tighten the restraits? > >>> > >>> 2. What is the meaning of the last column in torsion angle > >>> parameters: _chem_comp_tor.period, in cif file? In the PDB output > >>> file REFMAC also lists the RMS and WEIGHT for the torsion angles, > >>> period 1 through 4. > >>> > >>> 3. In REFMAC gui under Geometric parameters, there is > only one user > >>> controlled weight for torsion. By changing the weight > here, does it > >>> change the torsion weight for all 4 periods? > >>> > >>> Thanks in advance for the help. > >>> > >>> Huiying > >> > >> > >> > > > > > -- > Eckhard Hofmann <[EMAIL PROTECTED]> > Ruhr-Uni Bochum > AG Proteinkristallographie, LS Biophysik, ND04/316 > 44780 Bochum > Tel: +49-(0)234/32-24463, Sekr. -24461, FAX: -14762 > > 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 [EMAIL PROTECTED] 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
