On 15 Sep 2010, at 18:04, Ed Pozharski wrote: > On Wed, 2010-09-15 at 07:57 -0700, Pavel Afonine wrote: >> if you refined your structure with H, then you should deposit it with >> H > > sure. But the structure is not *refined with hydrogens* when they are > in predicted positions. Following the same logic one could suggest that > electron density should be deposited, since we can approximate it.
And I notice that a fair number of groups do deposit electron density - at least, they deposit PHIC and sometimes even HL coefficients in the sf.cif file. HL coefficients in the sf.cif file can get badly corrupted in the deposition process, but they definitely show willing. > I think it's useful to limit the information presented in a pdb-file to > what was actually refined + specific instructions on how the refinement > was done. I suppose I come to this from a background where every deposition is a fresh new test-case for new refinement software; it's only lack of download bandwidth and CPU power that makes me not want to start from the images. I like the idea that what you deposit is the output of a well-defined refinement; which means that you need to deposit the instructions for doing the refinement, and the model you used as input. There's a perfectly good PDB protocol for multi-MODEL files. Nobody does such depositions, I think the PDB would complain if you tried, and there's the problem of endless regression. I would be very happy if every PDB deposition with 'METHOD: MOLECULAR REPLACEMENT' had an extra MODEL in it containing the input to the molrep tool, and some REMARK lines describing how molrep was used; I would not complain if this was made compulsory for depositions which nowadays say 'STARTING MODEL: NULL'. 26 of the 130 depositions with method MOLECULAR REPLACEMENT this week have starting model NULL, as well as seven depositions with method FOURIER SYNTHESIS and starting model NULL. (why do MAD and SAD depositions still have a STARTING MODEL field?) (while we're on the subject of riding hydrogens, I would invite people to admire the conformations of the hydrogens in such places as the C-alpha of residues A45 and A57 of deposition 2x5n - it's clearly a software bug rather than any mistake on the part of the authors, but nonetheless striking) Tom
