I agree with Herman about personal preference but it also boils to our job as crystallographers to educate non-structural end-users. The fact of the matter is that a lot of researchers use structures without looking at the nuances of the PDB. It's actually pretty common among biologists to download a PDB file and build hypothesis or throw the coordinates into a downstream program like APBS or AutoDock without looking. They don't even realize that density exists or that they should check it, which makes the odds of them reading a header file for missing atoms or understanding the concept of b-factors and occupancy almost nil. Realizing that renders the argument moot until crystallographic data is demystified across the other sciences.
I will say that in principle I do like the idea of a data model + a projected model because it seems like something an end-user could wrap their head around, but in practice this would probably refuel the "what constitutes modelable density" debate all over again. Cheers, Katherine On Mon, Mar 26, 2012 at 10:19 AM, <herman.schreu...@sanofi.com> wrote: > Dear Ed, > > In the end it boils down to personal preferences. With the number of > crystal structures I refine each year, I am not going to go over every > flexible surface residue to decide whether to truncate the side chain or > whether there may be some low level density justifying to keep the side > chain, so I opt for the biochemical evidence. For me the added advantage is > that I only have a single pdb file to take care of. And again, I see no > problem in having a model with some atoms with a larger error bar. > > You are right that terminii are often truncated. In contrast to a missing > side chain, here we really have no reasonable hypothesis where the missing > residues are. They may even have been removed by a protease. > > Cheers, > Herman > > -----Original Message----- > From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Ed > Pozharski > Sent: Monday, March 26, 2012 4:50 PM > To: CCP4BB@JISCMAIL.AC.UK > Subject: Re: [ccp4bb] REFMAC5 residues with bad geometry > > On Mon, 2012-03-26 at 16:30 +0200, herman.schreu...@sanofi.com wrote: > > It is like with Heisenbergs uncertainty principle. Either one has a > > complete model with a number of atoms having a coordinate uncertainty > > of 4-6 Å, or one has a model where the uncertainty of all atoms is > > below say 0.5 Å, but with a lot of truncated side chains with clearly > > contradict available biochemical evidence. > > Excellent analogy. I am not sure why truncated arginine (as long as it is > not renamed to alanine) contradicts biochemical evidence though. > Termini are routinely truncated, no problem there. I have plenty of > biochemical evidence that there are more waters in the crystal than I model. > > If the truncated model contradicts biochemical evidence, the projected > model contradicts crystallographic evidence. I agree that a truncated model > may lead to interpretation problems, and thus the option of depositing a > projected model resolves that. > > Cheers, > > Ed. > > -- > Oh, suddenly throwing a giraffe into a volcano to make water is crazy? > Julian, King of Lemurs >
