[gmx-developers] Monte Carlo Simulations

[Anton Feenstra]

On 01/02/12 13:23, David van der Spoel wrote:
> On 2012-02-01 09:38, Anton Feenstra wrote:
>> Hi,
>>
>> Just plugging in our work, since I believe it will be relevant ;-)
>> (and my apologies if it turns out not so)
>>
>> A post-doc in our group, René Pool, has implemented a grand-canonical
>> Monte Carlo scheme using the GromPy gromacs-python interface. It is
>> based on library calls to the gromacs md engine from python, and
>> therefore can be very flexible. In our tests, it is just about as
>> efficient as a fully gromacs-internal implementation could be (i.e., the
>> overhead due to python etc, it only a few percent).
>>
>> There is an open source repo for GromPy
>> at gitub: https://github.com/GromPy
>>
>> There's a paper coming out in JCC soon, describing the interface and its
>> performance in calculating the critical properties of a LJ system and of
>> SPC water:
>> Enabling Grand-Canonical Monte Carlo: Extending the Flexibility of
>> GROMACS Through the GromPy Python Interface Module
>> René Pool, Jaap Heringa, Martin Hoefling, Roland Schulz, Jeremy C.
>> Smith, and K. Anton Feenstra
>>
>
> Dear Anton, this looks really useful. It would be even more useful if it
> could be employed with bleeding edge versions of gmx. In particular, I
> would be interested in running tons of small simulations and edit the
> tpr files in memory meanwhile. Do you think this kind of integration
> will be possible in the near future? I'm asking in particular since the
> gromacs you are using is somewhat dated (but for my purpose I would need
> the 4.6 development branch).

Yes, and we also really want to have it there. There are two things 
required:
1) small modification to the gromacs source, mainly the splitting of 
do_md into do_md_init, do_md_integrate and do_md_finalize; this is easy.
(Besides that, there is a list of memory clean-up to add so that 
do_md_integrate doesn't leak, also relatively easy.)
2) setting up the relevant data types into python (cpython IIRC). This 
is the painful part, but once done, it shouldn't need much re-doing. 
Unless, as I suspect may now be the case, gromacs' internal 
datastructure has changed. Someone should do that, but neither me nor 
René have the time at the moment. I'll try and get a good CS MSc student 
on it, I think. There may be some (semi-)automatic way to set this up, 
too, but that requires a bit of engineering (again, good for a CS MSc 
student...).


-- 
Groetjes,

Anton
  _____________ _______________________________________________________
|             |                                                       |
|  _   _  ___,| K. Anton Feenstra                                     |
| / \ / \'| | | IBIVU/Bioinformatics - Free University  Amsterdam     |
|(   |   )| | | De Boelelaan 1083A - 1081 HV Amsterdam - Netherlands  |
| \_/ \_/ | | | Tel +31 20 59 87783 - Fax +31 20 59 87653 - Room P136 |
|             | Feenstra at few.vu.nl - www.few.vu.nl/~feenstra/         |
|             | "Encountered Subspace Anomaly" (Star Trek)            |
|_____________|_______________________________________________________|