From: <rh...@fl...> - 2008-04-29 21:05:30
|
Bob Hanson wrote: > biomolecule oriented Jmol users: > > OK, I'm convinced by my colleagues here at St. Olaf and Carleton that > this is HUGE. Just playing with it a bit, I'm totally excited. You can > now load a subset of the atoms in a PDB file into Jmol (a limited SELECT > sort of filter to the LOAD command that works at the file-reading level > for PDB files) and then also select one of the BIOMOLECULE > specifications in the "REMARK 350" biological unit section of the header > and apply the symmetry described in that header. Bob, this is really great. When it works correctly it will help to reduce one major problem with large biological units: java memory. And it will also reduce the transfer volume from several megabytes to a few kilobytes. I tried revision 9339 from SVN with a few smaller examples (4otb, 1a34, 1sva) and they looked fine. At least they looked identical to the biological units provided by the wwPDB FTP site. In the case of '1a34' there are overlapping RNA chains but I guess this is because the symmetry operations are not defined properly. I also tried the currently most challenging entry '1m4x' (compressed file size: ~300 MB, uncompressed file size: ~1.3 GB). I was not able to automatically generate a picture of the biological unit with Jmol yet. (It hung up after several hours and using up about 30 GB of memory using Java 1.4.2_13, as far as I remember. Loading it interactively without creating an image did work.) So I used the "*.CA; BIOMOLECULE 1; APPLY SYMMETRY" filter to load the asymmetric unit file into the application and after ~5 minutes the structure appeared (using ~4GB memory for the whole JavaVM). But it looked quite strange, see the snapshot at http://www.fli-leibniz.de/~rhuehne/jmol/1m4x-biounit1-jmol-2008_04_29.png (colored by chain, not everything is visible) It should rather look as is shown at the RCSB page for '1m4x' (http://www.rcsb.org/pdb/explore/explore.do?structureId=1M4X). The file info in the menu showed the following: atoms : 2,081,520 bonds : 47,093 groups : 2,081,520 chains : 3 polymers: 1,950,530 The bond count seems to be wrong because some limit was reached according to the following message(s): BSPT tree depth too great:150 maximum auto bond count reached The polymer count is also wrong. Since there are 3 chains and 1680 units I think it should be 5,040. > I suspect as people play with this more, they will want me to expand on > that load filter select option. What else? Since we are automatically generating images of the biological units for all PDB entries with Jmol it would be very helpful if the filter would also be available as command-line option. I noticed that all generated copies are integrated into the same model. This prevents any easy way to select and display only specific copies (see also Mauricios answer). We have integrated the biological unit files available from the PDB now for about 2 years into the "JenaLib Jmol Viewer". In these files each copy is placed into its own model. This proved to be very useful, because it allows very easily individual selection and manipulation. Our viewer offers for example coloring each copy individually or coloring each chain individually (instead of by chain identifier). Regards, Rolf ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program. |