[gmx-developers] Re: CML and macromolecules

[Erik Lindahl]

Hi,

I strongly second David's comments; in Gromacs we want to use the files 
for two purposes:

1. Topology description of the system. CML works great for this, and I 
plan to completely get rid of our own topology file format, but then we 
probably need a parser that compares e.g. PDB files with a library of 
structures and constructs bonds that we know are correct instead of 
only using coordinate proximity. This is probably no big deal since we 
will anyway have to do it for changing termini, etc...

2. For structure prediction other non-simulation approaches we often 
need to do computations on hundreds of structures, but we want to 
retain every single piece of information in the PDB files we started 
from. Personally I would like to keep comments too, and even stupid 
numbering errors, so we can throw away the PDB files.


>> However I think there is still a strong "PDB-like" approach and I am 
>> happy
>> to extend CML to manage that aspect of macromolecules. I think it's 
>> *not*
>> useful for CML to try to model protein hierarchy
>> (primary/secondary/supersecondary/tertiary/quaternary, etc.) However 
>> it
>> could be useful to have a "flat-file" approach" where the atoms had 
>> PDB
>> like info on:
>> - their PDB type (CA, SG. etc.)
>> - their PDB number
>> - their residue type
>> - the chain number.
>>
>> CML could carry this - and more - , but would not support the explicit
>> hierarchy.

That would be great - I agree CML works best as a format for low-level 
chemical information and not include structural properties. As long as 
it is there in one way or another we can write it back to files if we 
really need to.

Cheers,

Erik