Re: [h5md-user] Additional properties of particles

[Jakub Krajniak]

Hi,

W dniu 07.11.2014 o 13:57, Pierre de Buyl pisze:

> > > 2. The 'resnr' that will behave like current 'species'. That
> > > property should group the particles in molecules. The .gro and
> > > .pdb file format have something similar. The residue name is not
> > > so important as it could be solved outside the h5md.
> > This point is less clear. The residual number does not affect the
> > physical properties/interactions of the particles (unlike "species"
> > does), its primary purpose is for analysis. (Please correct me if
> > I'm mistaken.)
> >
> > I'm not an expert in proteins, but the residuals seem to describe
> > the protein structure on a more coarse-grained level. So it has more
> > similarities with topology/connectivity (although it is less
> > stringent). What comes to mind is to store information about the
> > residual groups outside of the particles tree.
> >
> > Suggestions: store a list of particle indices (better: IDs) for each
> > residual. This makes it easy to look up all particles of given
> > residual, however the inverse would be inefficient.
> >
> > H5MD root
> >   \-- residuals
> >        \-- (name)
> >            \-- particle_subgroup1
> >            \-- particle_subgroup2
>
> As Konrad points out (other subthread), residue definition should be thought
> differently than what we have done so far. I don't know yet how it can be
> incorporated in H5MD as I don't really work with that kind of systems.
>
> But it maybe that Jakub (?) is interested in reproducing the feature that some
> MD packages have that each particle belongs (uniquely, then) to a molecule. This
> is not the straightforward translation of a physical property then.

Thanks for all suggestions. Indeed the residue id is not at all a particle 
property rather the description of the topology. I though about the concepts of 
subgroups but it seems a bit overhead. Let's say that I have 4000 chains of 
polyethylene so in the terms of subgroups I would have 4000 subgroups. On the 
other hand I could have a DataSet that will assign residue id to the particle id 
with very simple data structure even for time-dependent case.

I saw the MOSAIC but.. seems that the idea of fragments can also handle it.
The other solution in my case would be to use H5MD only as the particle position 
container and the topology information will be provided from external topology 
file. Shortly speaking I am trying to write H5MD reader for VOTCA 
(http://votca.org) package. It supports lots of formats but the main ideas 
inside the code are taken from GROMACS.

As for the charges that Konrad commented in other subthread I see the point. 
Perhaps MOSAIC can hold the data for charges.

Best,
Jakub

--
Jakub Krajniak, Computer Science Department @ KU Leuven
mob. +32 477 688 404
room: 200a 01.41

Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm