AMBER Archive (2008)
Subject: Re: AMBER: GB versus explicit medium
From: M. L. Dodson (activesitedynamics_at_comcast.net)
Date: Mon Nov 24 2008 - 09:43:52 CST
If you check the archives of the list, I think you will find postings
pointing out that simulating large (for some meaning of large) systems
with GB is actually computationally more expensive than using explicit
solvent. This is because of the large cutoff required for GB with the
accompanying very large number of pairs.
I'm thinking this was posted by Ross Walker (at least), but I am not
Francesco Pietra wrote:
> On Mon, Nov 24, 2008 at 12:08 PM, Carlos Simmerling
> <carlos.simmerling_at_gmail.com> wrote:
>> Using GB to replace explicit water already has some weaknesses, but I
>> think GB with dielectric of ~80 is an even worse model for the
>> interior of a lipid..
>> I'm not sure if that is the problem, but I would not really expect
>> this to work. Charlie Brooks has used his GB for membranes, but they
>> define a dielectric boundary at the water/lipid interface. There is a
>> review article on this in ARCC.
> I'll look for the article, it would help if you have the exact ref.
> Actually, my initial plan was to embed the pore region into a hydrated
> POPC membrane and define GB for the extracellular part. I first tried
> GB for the whole, getting discouraged by the high computational cost.
> Adding the membrane it would be to much, or too costly. Finally, given
> the nature of the protein, with so many variables for the status of
> the amino acids (in the GB simulation the protein has a charge -50!
> and the best pKa simulation programs are defeated by so much stuff: I
> tried) this is a project for the future, or for unlimited
> computational resources. Unless Charlie Brooks' methodology is (for
> me) computationally affordable, I am forced to to leave docking to the
> extracellular part at the status that the docking program was able to
> do; it works through a huge number of reciprocal orientations but the
> proteins are left to their crystallographic conformations. Certainly
> not the best model since the crystallographic data are at a very low
> temperature. I would be happy enough to be able to pave the way for
> future research.
> I agree that there may be a bias overriding the capability of GB. I
> tried to be careful with the steps preceding production, following
> advice that you kindly provided before (notice that there was no
> protein distortions before MD production; distortion - as described -
> came out slowly and gradually during production). How production is
> carried out in steps was outlined in my previous mail.
>> in order to make the GB vs explicit comparison that you did below, you
>> really need to compare the same system, not a comparison where ones
>> has many more restraints.
>> On Mon, Nov 24, 2008 at 4:39 AM, Francesco Pietra <chiendarret_at_gmail.com> wrote:
>>> Hi all:
>>> Is any experience in carrying out long MD simulations for
>>> transmembrane proteins under both GB conditions and explicit
>>> This question because I am carrying out MD with a large transmembrane
>>> protein, composed of several identical units and carrying a docked
>>> small protein (docked in accordance with physiological data) in the
>>> extracellular part. MD under GB conditions after a careful
>>> minimization and heating to 300K. Force field ff99SB.
>>> MD under GB at 300K
>>> with restraint on Cl-
>>> using pmemd in amber10
>>> imin=0, irest=1, ntx=5, ntb=0,
>>> ig=-1, igb=5, nrespa=2, ntc=2,
>>> ntf=2, ntt=3, gamma_ln=2.0,
>>> nstlim=30000, dt=0.002,
>>> ntpr=100, ntwx=100,
>>> tempi=300.0, temp0=300.0,
>>> cut=999.0, rgbmax=999.0,
>>> Keep Cl- 427 restrained
>>> RES 427
>>> I understand that, forced by the huge system, I am using rather
>>> drastic conditions (dt=0.002, nrespa=2), though I was badly surprised
>>> to see the initial portion of one of the chains of the pore region
>>> undergoing a 90 degrees bent at SER 14 (14 from the beginning of the
>>> chain, while the chain continues for ca 23 amino acids in the pore
>>> region). EPTOT EKTOT and ETOT remain constant. The extracellular
>>> portion and docked small protein also undergo some conformational
>>> adjustment, though, no "distortion" of the system occurs.
>>> In another approach, for different reasons, I isolated the pore region
>>> (with a small molecule docked in; docked in accordance with
>>> physiological data)), cutting away the extracellular portion and
>>> restraining the ends of the chains at the end of the pore region.
>>> Under periodic conditions, in a hydrated POPC membrane, for a much
>>> longer MD than the above, no distortion of the pore region occurred
>>> Production MD, restraining
>>> capping groups
>>> imin=0, irest=1, ntx=5,
>>> nstlim=100000, dt=0.002,
>>> cut=10, ntb=2, ntp=1, taup=2.0,
>>> ntc=2, ntf=2,
>>> ntpr=1000, ntwx=1000,
>>> ntt=3, gamma_ln=2.0,
>>> Keep ACE 79 restrained
>>> RES 79
>>> many other capping groups restrained.
>>> I wonder whether this may be a limitation of GB or there is evidence,
>>> or suspicion, that setting up was not adequate. Surely, if the GB
>>> simulation has to be made some fold slower that it is, the simulation
>>> could hardly be carried out for an adequate time.
>>> francesco pietra
M. L. Dodson
Business Email: activesitedynamics-at-comcast-dot-net
Personal Email: mldodson-at-comcast-dot-net
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)