On Oct 8, 2009, at 8:35 PM, Dina Schneidman wrote:
That is what I always assumed :-)Protein is more than a chain. Chain corresponds to tertiary structure.
Protein's quaternary structure can have more than one chain!
A classic example is hemoglobin, 4 chains. Another classics is
antibody, 2 chains.
No one said anything about getting rid of chains. Just about getting rid the PROTEIN/CHAIN distinction.
So we need chains around! and also how can we add bonds without
chains? do you plan to connect them together?
It does too define a hierarchy:
and let me put two more cents:
PDB format does not define any hierarchy. it is a set of atoms.
a pdb file contains models
models contain chains and heterogens
chains contain residues
residues contain atoms
heterogens contain atomsI don't see that the second follows at all from the first. In fact I would say quite the opposite. But I disagree with the first, so it doesn't really matter :-)
if we
want to build an hierarchy out of PDB it should clearly follow from
the format.
Plus the parallel various ligands and stuff which also need to get attached to root (preventing root from being a protein or molecule).
So the best way is to have 4 levels that are well defined
by the corresponding PDB fields:
Atom, Residue, Chain, Root
I think all other assumptions are only assumptions and a good source for bugs.
To put one of the problems another way, the big problem is that, ultimately, one would like a hieararchy with a molecule (protein) containing multiple chains.
The PDB reader can't, in general, create such a thing since it doesn't know how the chains are grouped into molecules. As a result, it has to return something intermediate.
Currently it returns a hierarchy would would have to be broken apart and put back tother in order to get the presumably desired result.
I would suggest producing a vector of molecules instead so that the user can filter them/assemble them as needed.
We could provide special versions of the reader to handle easy cases (like where you know the pdb file only contains one protein).
_______________________________________________
IMP-dev mailing list
" target="_blank">
https://salilab.org/mailman/listinfo/imp-dev