Probably the most confusing aspect of the loading/model creation mechanism. See John Beaver's
Ids stored in the model are
ids, not Pdb. Pdb ids are looked up.
One problem is the conversion methods are quite hard to use - they return a two-element array of
objects which have to be tested for existence and cast. I'm currently working on providing
Ndb ids primarily come from .cif/.xml files, Pdb ids from .pdb files. The identification schemes are quite different.
Thus, the requirement to map from one to the other. The PdbToNdbConverter performs this conversion.
- On loading XML files, the chain and residue ids are extracted in both Ndb and Pdb namespaces.
- On loading PDB files, the Ndb ids are set to their corresponding Pdb ids, thus the mapping is essentially 1:1.
The loaders create the PdbToNdbConverter as the last step from the lists of names extracted. It is handed off to the StructureMap, which then uses it throughout the rest of the application.
Non-protein chains present their own issues -
From John Beaver (edited):
Pdb and Ndb deal with one of the major legacy problems of the PDB data.
The old .pdb file format has been around for a very long time. It’s simple, and it’s what most people who don’t use the website use. It has several technical limitations, but the data matches the original author submission very closely.
This is a problem. Very commonly, a small molecule or DNA strand will have the same chain ID as a protein chain, for example. This can cause problems when the viewer is deciding where to draw ribbons and bonds.
The Ndb (whose name I took from one of the Xml tags in the PDB XML format and which may or may not be proper terminology) is a separate namespace for chain IDs and residue IDs. It is much more highly cleaned; you’ll almost never see a small molecule or DNA chain mixed with protein in one chain. Also, PDB residue IDs can have letters in them; NDB residue IDs are always integers.
The Ndb namespace still has data cleanliness problems, but it seems much better overall than the Pdb namespace.
For an example of what I mean, look at the following .xml snippet.Scroll about halfway down the file, and you’ll see something like…
<PDBx:label_alt_id xsi:nil="true" /><br/>
<PDBx:label_asym_id>A</PDBx:label_asym_id> (--> NDB chain ID)<br/>
<PDBx:label_seq_id>165</PDBx:label_seq_id> (--> NDB residue ID)<br/>
<PDBx:auth_seq_id>165</PDBx:auth_seq_id> (--> PDB residue ID)<br/>
<PDBx:auth_asym_id>E</PDBx:auth_asym_id> (--> PDB chain ID)<br/>
Here, label_asym_id is the NDB chain ID and auth_asym_id is the PDB chain ID. Similarly, label_seq_id is the NDB residue ID and auth_seq_id is the PDB residue ID.
To make matters worse, Phil Bourne insisted that the community prefers to see the PDB nomenclature. This is correct, since most of the community uses the .pdb format. Whereas the NDB nomenclature is *much* more amenable to use in the internal data structures, I had to make a large dictionary to translate NDB to PDB