AMBER Archive (2009)

Subject: Re: [AMBER] PMEMD vs GROMACS?

From: Sasha Buzko (obuzko_at_ucla.edu)
Date: Tue Feb 24 2009 - 11:15:59 CST


Thank you for the detailed explanation, Vlad.
Since we are working with fairly simple soluble proteins, we could
probably give a try to several options. My guess is that this type of
system is relatively well worked out in most of the popular force fields..
And since GROMACS does support Amber force field parameters (comment by
Hannes). we'll probably check it out and compare to PMEMD that we use
right now.

Thanks again

Sasha

Vlad Cojocaru wrote:
> Hi Sasha,
>
> Well .. Its a good question what you are asking ... There are however
> other MD software which are widely used such as NAMD, CHARMM and so on
> ...
> GROMACS 4.0 seems to be fast indeed. However, on our small cluster
> here I have NAMD and PMEMD scaling up to 64 cores while GROMACS scales
> only until 16 or 32 depending on the system ..
>
> But the main difference between these programs is the force fields you
> can use with them (and make sure not to mix the names of the programs
> with the names of the force fields, that's why I use upper case for
> programs, lower case for force field names). AMBER is designed to work
> with the force fields based on Cornell et al. 94 (amber force fields),
> GROMACS works with opls and gromos force fields, NAMD works mainly
> with charmm. In NAMD you can also use the amber force fields if you
> follow strictly the recommendations. In GROMACS you cannot use neither
> charmm, nor amber force fields unless you implement it yourself.
>
> There are also slight differences in the options provided by these
> programs. One typical issue is the pressure coupling scheme for NPT
> simulations. GROMACS and NAMD provide more options here while AMBER
> only provides Berendsen's coupling algorithm. So, if you run membrane
> simulations, AMBER is still limited. In return, AMBER works with the
> glycam force filed for sugars. So, if you have a glycosylated protein,
> then probably you'd like to use the glycam force field to treat the
> sugar(s).
>
> At the end, it comes down to which force field you'd like to use. And
> this is dependent on the problem you are investigating. If you have
> membranes, even if you'll use the amber force fields (e.g gaff), you
> might want to use the NAMD program for the runs as it allows more
> sophisticated pressure coupling algorithms. If you have only soluble
> proteins, it probably depends very much what specific issue you are
> investigated. All these force fields seem do well on some protein
> issues and not well on others. If you have nucleic acids, the amber
> force fields are probably the most widely used, although the charmm
> force fields have also been successfully applied.
>
> Only after deciding the force field you want to use, it comes down to
> the program you want to use depending on the architecture you have.
> NAMD is well known to do very well on scaling up to thousands of cores
> on big computers such as Blue Genes. GROMACS is still lagging behind
> in terms of scaling although its very fast. With PMEMD, I do not have
> experience on big machines, but PMEMD is faster comparing to NAMD on
> up to 64 cores on opteron clusters. Unfortunately, PMEMD has still
> limited functionality. For methods such as SMD, Umbrella Sampling,
> Targeted MD you still need to use SANDER, and this is much slower and
> scales much worse. That's why I always use NAMD with the amber force
> fields if I need to do such calculations.
>
> After this, you need to read literature on your specific topic to see
> what to use and how ...
>
> Cheers
> vlad
>
> Sasha Buzko wrote:
>> Hi all,
>> I wonder if anyone has had any comparative experience with pmemd and
>> gromacs. The latter seems to be faster, but what about the simulation
>> outcomes? Are the results identical/similar and if there are
>> differences, how significant?
>>
>> And if the results are comparable, a more general question: why
>> doesn't everyone just switch to using gromacs? Are there other issues
>> that come into play?
>>
>> Thanks for any insight.
>>
>> Sasha
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER_at_ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber