AMBER Archive (2009)

Subject: Re: [AMBER] Problem with entropy calculation of bigger system using NMODE and NAB.

From: Marek Malý (maly_at_sci.ujep.cz)
Date: Wed Jan 21 2009 - 16:48:58 CST


Dear professor Case,

thank you again for the answer. I apoligize for the confusion regarding to
"frms memory dependence". I thaught that was pretty clear that I wrote
to this forum due to NMODE/NAB memory allocation problems and not because
of eventual unrealistic results based on failure in parameters setting
(here underestimation of number of CG and NR minim. steps).
Of course that enormously small number of minimisation steps in my input
file was not just the proof of
my enormous ignorance but it was rather just the precaution which should
prevent me from big time losses, in
my NMODE/NAB memory experiments with 15000 atoms structure. Naturally that
I assumed that you and Niel are aware of this fact and
that's why I immediately derived from your answers "frms memory
dependence" :))

OK, lets go back to my problem.

I would like to calculate entropy of big system (15000 atoms) with NAB
which requires a little more
than 17GB of RAM memory and I dispose of cluster with 8CPUs nodes - 16GB
of shared memory per node.

I see this possible soulutions:

A)
After asking my collegues to leave one entire node empty just for me I
could find,
that 16GB is enough since system is able to use for the rest requested
memory (cca 1GB)
swap disk. Although this could be pretty slow solution , thanks God for it
!

B)
Some hope provide me the fact that NAB could be now implemented also in
parallel version.
If I run NAB on 2 empty nodes, I have 2x16GB of shared memory which could
solve my problem,
of course only in the case that each node do not require copy of the whole
dataset. I cannot
verify this now since in this moment I have only serial version installed,
so eventual
preliminary comments from the erudite person are welcomed !

C)
The last possibility in this story is probably calculation of the entire
system using some "per partes" approach
or to approximate the whole entropy by some most important contribution
("partial entropy").
To be allowed go this way some restriction tools (restraint-masks like in
sander MD or minimisation) should be available
also for normal-mode analysis. Am I right ? If yes is something like this
recently available in NAB ?

I also thanks a lot for the answer regarding to mm_PBSA x NAB discussion.
I have just a small
additional question.

If I use recommended approach (= enthalpic contribution with mm_PBSA where
for E_desolv use PB model and NAB use only for the
calculation of the entropic contribution) it is probably appropriate to
use in NAB nonzero value for gb although previous calculation
was carried out using PB model. Still better to "mix" two solvation models
than solvation model and situation in vacuum (gb=0).
Am I right ? Regarding to available nonzero gb values are there any
general recommendations rigarding possible values 1,2,5 ?
For example if I tell you "I am simulating big dendrimer (spherical shape)
with 21 base-pair siRNA, where strongly dominate
electrostatic interaction (positively charged dendrimer and negatively
charged siRNA)" is it clear which gb model could be
the most suitable for this case ?

   Thanks a lot in advance for the answers (hopefuly last in this topic :))
)

              Marek

ss
Dne Wed, 21 Jan 2009 15:52:33 +0100 David A. Case
<case_at_biomaps.rutgers.edu> napsal/-a:

> On Wed, Jan 21, 2009, Marek Mal? wrote:
>>
>> From your responses clearly follow that memory requirements (for N-R
>> minimisation and normal mode analysis) are strongly
>> dependent on quality of the previous conjug. grad. minimistation or more
>> precisely on the size of the "frms".
>
> No: the memory requirements are independent of frms. It is just that NR
> minimization (as coded here) requires a good starting point.
>
>> OK in mm_PBSA is possible to choose between PB and GB algorithms to
>> calculate desolvation energy. I always used PB
>> since I read somewhere that it is let say more realistic model but if I
>> am not wrong in NAB is available only GB
>> in this moment so what is the better choice (from the point of view of
>> more realistic results).
>
> PB is generally preferred for "snapshot" analysis. The recommendation
> for NAB is only for the normal mode calculations -- other parts of the
> calcualations can use other programs.
>>
>>
>> #1 - to use mm_PBSA for the enthalpic contribution with PB method to
>> calculate desolvation energy and
>> than to use NAB with gb=1 to calculate entropic contribution.
>
> The above is what I recommend. No matter what you do, the entropy part
> will be dicey.
>
> ...dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> __________ Informace od NOD32 3785 (20090121) __________
>
> Tato zprava byla proverena antivirovym systemem NOD32.
> http://www.nod32.cz
>
>

-- 
Tato zpráva byla vytvořena převratným poštovním klientem Opery:  
http://www.opera.com/mail/

_______________________________________________ AMBER mailing list AMBER_at_ambermd.org http://lists.ambermd.org/mailman/listinfo/amber