Is n't the danger here that you are just fitting to noise generated by the rest of the Unit Cell that you are ignoring. So it is easy to do a 'good bit' of modelling in a specific region. Just like some tennis players have very good serves but do not win matches.
________________________________ From: George Devaniranjan <devaniran...@gmail.com> To: CCP4BB@JISCMAIL.AC.UK Sent: Wednesday, 11 June 2014, 19:22 Subject: Re: [ccp4bb] EDSTATS for an extracted fragment Thank you Gerard for your suggestion. I will look into the paper you suggested and MAPMAN too. I am "idealizing" fragments of PDB and while this would not agree when compared with the well refined structure, my goal really is seeing if these "ideal" fragments can still be identifiable when I look at the electron density and how to quantify that. (Ideal here refers to a small number of Phi and Psi values I picked as "representative" of different regions in the Ramachandran map). Isn't the danger here that you are can't tell how much improvement is just fitting to noise from improvements that should be made in the rest of the model? Unless you go on and do a final full refinement you can't know if your modelling is warranted by a global improvement in fit. It could be the case that you are improving globally but I don't know how you can tell that by fitting to just a bit of the map. We have all had that experience of rebuilding a residue so it looks 'right' to us locally but then the refinement program knows better after making a global assessment and puts it back. If some way to give local density 'error bars' that might be helpful so that a maximum likelihood approach could be used during real space fitting. Is that similar to what those useful LLG maps give in heavy atom model refinement? regards Martyn Cambridge I picked on EDSTATS to explore since it gave individual residue information but realize now that since I am dealing with a small sub-set of residues an alternative is required. On Wed, Jun 11, 2014 at 2:01 PM, Gerard DVD Kleywegt <ger...@xray.bmc.uu.se> wrote: What you want is a test for how well each model agrees with its own map. It is fair to argue that the model that is more self-consistent (agrees better with its own map) is the better model. But you won't learn that by comparing model A to map B. >> > However, conversely, if your modified model fits the original map better than the model that was used to calculate the map itself, you've done a good bit of model building. If you want to do this calculation (with all the warnings and caveats), you can also use MAPMAN - http://xray.bmc.uu.se/usf/mapman_man.html#S41 . The method you propose is essentially the same as this one: http://www.ncbi.nlm.nih.gov/pubmed/18598022 but for a fragment of your macromolecule instead of for a ligand (if you don't have access to the journal, you can request a reprint here: http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=87 ) > >--Gerard > >****************************************************************** > Gerard J. Kleywegt > > http://xray.bmc.uu.se/gerard mailto:ger...@xray.bmc.uu.se >****************************************************************** > The opinions in this message are fictional. Any similarity > to actual opinions, living or dead, is purely coincidental. >****************************************************************** > Little known gastromathematical curiosity: let "z" be the > radius and "a" the thickness of a pizza. Then the volume > of that pizza is equal to pi*z*z*a ! >****************************************************************** >
