[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [h5md-user] New member
From: |
Pierre de Buyl |
Subject: |
Re: [h5md-user] New member |
Date: |
Thu, 1 Sep 2011 22:04:22 +0200 |
Hello Konrad,
Thanks for joining us and providing your input.
As you could understand, we try to standardize first a core set of
data elements and proceed then based on accumulated experience.
Regards,
Pierre
Le 1 sept. 11 à 19:56, Konrad Hinsen a écrit :
Hello everyone,
I just joined the group after some discussions with Pierre about
HDF5 trajectories at the recent EuroSciPy conference in Paris.
A few words about my background and interest. 15 years ago, I
started to write a library for biomolecular simulations in Python
(with some C code for speed). It's called the Molecular Modelling
Toolkit and you can get it from
http://dirac.cnrs-orleans.fr/MMTK/
At the time, I spent some time thinking about trajectory storage
and I ended up choosing the netCDF format/library. HDF5 wasn't out
yet, and HDF4 was much less interesting. Now we want to migrate to
HDF5 for a variety of reasons, the most important ones being the
parallel I/O option and the possibility to combine many datasets in
a single file using HDF's hierarchical structure. So the h5md
initiative comes just at the right moment.
In one aspect we are much more ambitious than h5md: we want to be
able to store the full description of a molecular system, including
the chemical structure of all molecules, the nature of the atoms or
pseudoatoms, the rules for calculating the potential energy, and
various other parameters. Since that is the hardest part, we
started at that end, and by now I have a fairly useable data model
for molecular systems with a Python in-memory representation and an
on-disk HDF5 representation. I am currently doing extensive tests,
such as converting a large subset of the Protein Data Bank (PDB) in
order to see of everything in there is fully representable. I plan
to make a first version public once those tests are done. What my
molecular data model defines is how molecules map to atom indices
in configurations and trajectories, so it should in principle fit
in well with h5md, which I will of course explore.
Konrad.
--
---------------------------------------------------------------------
Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: research AT khinsen DOT fastmail DOT net
---------------------------------------------------------------------
-----------------------------------------------------------
Pierre de Buyl
Chemical Physics Theory Group - University of Toronto
Physique des Systèmes Complexes et Mécanique Statistique - Université
Libre de Bruxelles
web: http://homepages.ulb.ac.be/~pdebuyl/
-----------------------------------------------------------