Hello,
The main issue that Gustavo is talking about is the necessary value for
cut. With periodic boundary calculations, long-range electrostatic
interactions are calculated via PME, so "cut" can remain small (~8 Å)
without introducing error. This is because the only term really being
truncated at this distance are VDW, which are r^-6, so the magnitude of
these terms drops off significantly. However, electrostatic interactions
are too strong to truncate at 8 Å, so if you aren't using pme (which
implicit solvent doesn't use), then you need to drastically increase cut so
as not to miss a large part of the electrostatic energy. However, such a
large increase in cut causes the non-bonded list and the non-bonded
calculation to take an extremely long amount of time, possibly even more
expensive than explicit solvent.
Also, 15K atoms doesn't seem THAT large, and seems quite feasible with a
parallel executable (not sure how many waters are added...) You may just
have to get access to some supercomputer time.
Hope this helps,
Jason
On Fri, Jul 9, 2010 at 9:45 AM, Yan Gao <yan.gao.2001.gmail.com> wrote:
> Hi Gustavo,
>
> Thank you for your comment. It is a very important issue.
> I have not run any simulation with amber, but just through an example from
> the mailing list.
> I would be very grateful if you can give me further suggestion and comment.
>
> Yes my system contains 10 ~ 15k atom, ~5k for not quite flexible carbon
> nanotubes, others for flexible molecules.
> Gromacs indicates that I need 6 years to run 1ns simulation without
> parallel
> run.
> A post by a user in amber said he/she simulate 15k atom with implicit
> solvent, and it will be ~60 days for 1 ns without parallel run.
> I assume that:
> 1: I can use amber with implicit solvent and use parallel run.
> 2: ~5k CNT carbons are relatively rigid and may not require much
> computation
> time for surface geometry calculation in implicit solvent case.
> 3. The parallel I can access gives ~15X for either amber or gromacs.
>
> Any input is welcome!
>
> Yan
>
> On Fri, Jul 9, 2010 at 3:50 AM, Gustavo Seabra <gustavo.seabra.gmail.com
> >wrote:
>
> > Are you sure you *need* implicit solvent for this calculation?
> >
> > For large systems, implicit solvent calculations can get slow,
> > sometimes even slower than explicit solvent ones.
> >
> > Gustavo.
> >
> >
> >
> > On Fri, Jul 9, 2010 at 1:51 AM, Yan Gao <yan.gao.2001.gmail.com> wrote:
> > > Thanks Jason.
> > >
> > > Yes I also feel that would be tricky. But The carbon nanotube has quite
> > > tricky structure that if you break a tube into two for topology purpose
> > > (with antechamber), when putting their topology together, I feel that I
> > need
> > > to manually add a lot of bond/angles/dihydrals. And I may need to use
> > > different sized nanotubes.
> > >
> > > As for why I changed from gromacs to amber. I want to simulate ~ 15k
> > atoms
> > > containing nanotubes and polymers. It may take gromacs >100 years for 1
> > ns
> > > with explicit water.
> > > And I was told that amber has the best implicit solvent simulation
> > ability.
> > >
> > > BTW. It seems that gromacs has coarse-grain ability but no implicit
> > solvent,
> > > while amber has implicit solvent but no coarse-grain.
> > >
> > > Yan
> > >
> > > On Thu, Jul 8, 2010 at 8:26 PM, Jason Swails <jason.swails.gmail.com>
> > wrote:
> > >
> > >> On Thu, Jul 8, 2010 at 7:26 PM, Yan Gao <yan.gao.2001.gmail.com>
> wrote:
> > >>
> > >> > Thanks Bill.
> > >> >
> > >> > I am going to take your advice and try to write a program to convert
> > the
> > >> > gromacs topology to amber's.
> > >> >
> > >>
> > >> This could be quite tricky. In order to do this, you'll have to
> > understand
> > >> both types of topology files as well as how to convert one into the
> > other.
> > >> Understanding the Amber topology file itself is very difficult. You'd
> > >> probably be better off using pre-established programs to create your
> > >> topology file(s).
> > >>
> > >> I have another question regarding the force field.
> > >> > I use customized FF in gromacs, so I would wish to keep it
> customized
> > in
> > >> > amber. May I know if it is doable? And if a customized FF will
> > physically
> > >> > conflicted with the implicit solvent which I want to apply? (Or if
> the
> > >> > result be un-trustable ) Thanks.
> > >> >
> > >>
> > >> Why are you trying to switch to Amber? Is there something you can do
> > with
> > >> Amber that you can't with gromacs?
> > >>
> > >>
> > >> Good luck!
> > >> Jason
> > >>
> > >>
> > >> > I am also will include free ions in the system with implicit
> solvent.
> > >> Will
> > >> > the free ions be an issue? Thanks.
> > >> >
> > >> > Regards,
> > >> > Yan
> > >> >
> > >> > On Thu, Jul 8, 2010 at 2:28 PM, Bill Ross <ross.cgl.ucsf.edu>
> wrote:
> > >> >
> > >> > > > I want to simulate carbon nanotubes (CNT) with implicit water,
> so
> > I
> > >> > need
> > >> > > to
> > >> > > > translate the topology from Gromacs format to Amber format.
> > >> > >
> > >> > > > I tried antechamber using the pdb file, planning to manually
> > modify
> > >> the
> > >> > > > parameters after the file is constructed. However my tube is big
> > >> (1200
> > >> > > > carbon atoms), and antechamber said "The atom number exceeds the
> > >> > MAXATOM,
> > >> > > > reallocate memory", and eventually crashed.
> > >> > >
> > >> > > > I also think of divide the long tube into small ones, but it
> seems
> > >> that
> > >> > I
> > >> > > > need to manually add the bond/angle/dihydral to connect the
> small
> > >> > tubes.
> > >> > > The
> > >> > > > number of connections seems to be quite large, and I am not
> quite
> > >> know
> > >> > of
> > >> > > > how to add it. So I still look for a better way of doing it.
> > >> > >
> > >> > > You will need bond, angle and dihedral parameters whether the
> > >> > > molecule is one residue or many. But specifying the many
> > >> > > connections between multiple smaller residues could be a chore.
> > >> > > You might want to write a program to generate the leap cmds
> > >> > > to do this. It might also be tricky to turn a 1-residue pdb
> > >> > > into a multi-residue pdb; if the 1-res pdb was generated by
> > >> > > a program, you may want to modify that program.
> > >> > >
> > >> > > Given all that, it may be worth trying to debug antechamber
> > >> > > to allow the 1-res version.
> > >> > >
> > >> > > Bill
> > >> > >
> > >> > > _______________________________________________
> > >> > > AMBER mailing list
> > >> > > AMBER.ambermd.org
> > >> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >> > >
> > >> >
> > >> >
> > >> >
> > >> > --
> > >> > Yan Gao
> > >> > Jacobs School of Engineering
> > >> > University of California, San Diego
> > >> > Tel: 858-952-2308
> > >> > Email: Yan.Gao.2001.gmail.com
> > >> > _______________________________________________
> > >> > AMBER mailing list
> > >> > AMBER.ambermd.org
> > >> > http://lists.ambermd.org/mailman/listinfo/amber
> > >> >
> > >>
> > >>
> > >>
> > >> --
> > >> Jason M. Swails
> > >> Quantum Theory Project,
> > >> University of Florida
> > >> Ph.D. Graduate Student
> > >> 352-392-4032
> > >> _______________________________________________
> > >> AMBER mailing list
> > >> AMBER.ambermd.org
> > >> http://lists.ambermd.org/mailman/listinfo/amber
> > >>
> > >
> > >
> > >
> > > --
> > > Yan Gao
> > > Jacobs School of Engineering
> > > University of California, San Diego
> > > Tel: 858-952-2308
> > > Email: Yan.Gao.2001.gmail.com
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Yan Gao
> Jacobs School of Engineering
> University of California, San Diego
> Tel: 858-952-2308
> Email: Yan.Gao.2001.gmail.com
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Graduate Student
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jul 09 2010 - 08:30:04 PDT
