That was an awesome explanation for the impropers (I should have thought
about bonds angles taking precedence) thanks. This fact has not been
adressed properly anywhere atleast as far I have searched. I couldn't find
this in cornell's paper or GAFF (junmei) or may be it is obvious for people
who have experience in MD.
I rely on parmed.py extensively for my simulations, often to check the
dihedral connectivity information and also for exclusions.
Your statment :
"However, since the
atom types defined in that torsion *may* be used elsewhere in your
molecule, you need to make sure that the nonbonded scaling factors for that
parameter is "correct".
To take into consideration the above statement, in my simulation I have the
number of dihedral types equal to number of dihedrals that is each
dihedral has its own type (as defined by parmed.py). Although I am afraid
that the end group interactions (1-4 non bonded) of a multiterm dihedral
(atoms with the same type ) may be counted twice ( I do not know yet how
lammps calculates the 1-4 interaction for a multi term dihedral, if the
multiterm dihedral is defined twice as two new type of dihedrals)
Your opinion on my above imeplementation would be helpful
Thanks
Arun
On Wed, Oct 15, 2014 at 8:51 PM, Jason Swails <jason.swails.gmail.com>
wrote:
> On Wed, Oct 15, 2014 at 8:05 PM, Arun Srikanth <askforarun.gmail.com>
> wrote:
>
> > Thanks Daniel. I understand that the sum is for all the four terms of the
> > dihedrals.
> >
> > For example lets say that I have the following dihedrals which I obtained
> > from parmed.py.
> >
> > From your reply I understand that the sum is over 9 dihedrals and I also
> do
> > not have to count the 1-4 interaction for the last five dihedrals but if
> > you see the vdw and eel weighting factors from parmed it is not zero for
> > the fifth dihedral term (last two terms) in which M is printed. It is
> zero
> > for the the last four dihedral term as you say, which I am fine with. Why
> > does parmed.py does not print zero for the fifth dihedral
> >
> > 1 C ( c3) 9 C2 ( c3) 11 C3 ( c3) 12
> > O1 ( os) 0.1556 3.0000 0.0000 1.2000 2.0000
> > 5 C1 ( c3) 30 C20 ( c3) 29 C19 ( c3) 28
> > O2 ( os) 0.1556 3.0000 0.0000 1.2000 2.0000
> > 9 C2 ( c3) 11 C3 ( c3) 12 O1 ( os) 13
> > C4 ( ca) 0.3833 3.0000 0.0000 1.2000 2.0000
> > 10 O ( oh) 9 C2 ( c3) 11 C3 ( c3) 12
> > O1 ( os) 1.1750 2.0000 0.0000 1.2000 2.0000
> > M 10 O ( oh) 9 C2 ( c3) 11 C3 ( c3) 12
> O1
> > ( os) 0.1440 3.0000 0.0000 1.2000 2.0000
> > I 13 C4 ( ca) 15 C6 ( ca) 14 C5 ( ca) 36
> > H10 ( ha) 1.1000 2.0000 180.0001 0.0000 0.0000
> > I 14 C5 ( ca) 16 C7 ( ca) 15 C6 ( ca) 37
> > H11 ( ha) 1.1000 2.0000 180.0001 0.0000 0.0000
> > I 16 C7 ( ca) 18 C9 ( ca) 17 C8 ( ca) 38
> > H12 ( ha) 1.1000 2.0000 180.0001 0.0000 0.0000
> > I 13 C4 ( ca) 17 C8 ( ca) 18 C9 ( ca) 39
> > H13 ( ha) 1.1000 2.0000 180.0001 0.0000 0.000
> >
>
> All ParmEd prints is *exactly* what is in the prmtop file. If there is an
> M or I next to it (as ParmEd prints it -- E or I as cpptraj/rdparm prints
> it), you do not count those 1-4 nonbonded interactions. Period. The
> scaling factors become irrelevant. I see what is confusing you, but this
> is simply an implementation detail that you can only know if you understand
> the history behind the scaling factors defined in the prmtop file or have
> heard somebody that knows this history describe it before. I've given my
> explanation in the postscript, since it is long. [1]
>
>
>
> > 2) Which part of the paper discusses that 1-4 interactions of the
> improper
> > should not be included ? ? Can you please guide me to the paper or
> journal
> >
> > which says this.
> >
>
> Look at the definition of an improper torsion. It involves a central atom
> that is bonded to every *other* atom. As a result, every atom in the
> improper forms either a bond *or* an angle with every other atom in the
> improper. And bonds and angles are completely excluded by the Amber force
> field. This isn't a case of impropers being treated "specially" -- it's a
> case of the bond- and angle- exclusions taking precedence (as they always
> do).
>
> HTH,
> Jason
>
> [1] So the scaling factors are the quantities by which the 1-4 nonbonded
> terms are *divided* to get the actual interaction. When the scaling
> constant is set to 0, this term becomes infinite. There are some torsion
> terms defined in the dihedral whose end groups are ALWAYS excluded --
> specifically the first N-1 terms of an multi-term torsion with N terms (so
> you only count the 1-4 nonbonded interactions for the N'th term). If you
> set these scaling factors to 0 and some code actually (erroneously)
> computes the 1-4 nonbonded interactions for that torsion term, your energy
> and forces will become infinite. It's a way of making sure that a buggy
> program blows up rather than spits out reasonable (but still very *wrong*)
> answers.
>
> However, there are certain torsion parameter types for which the end groups
> are excluded in some specific torsions, but are not excluded for other
> torsions. This happens specifically in systems containing a 4-, 5-, or
> 6-membered ring. For the purpose of this explanation, only consider atoms
> _in_ the ring. In a 4-membered ring, atoms that are 1-4 to each other are
> also bonded to each other. In 5-membered rings, atoms that are 1-4 are
> also angled to each other. In 6-membered rings, atoms that are 1-4 to each
> other are 1-4 to each other through a *different* set of 3 bonds. Draw out
> examples to convince yourself that what I said was true. In this case,
> some of the torsions defined here must be defined as "M" (or "E") to make
> sure that their "end-group interactions" (i.e., 1-4 nonbonded terms) are
> omitted, since they are either already excluded by bond or angle terms, or
> they're already included by *another* torsion term. However, since the 4
> atom types defined in that torsion *may* be used elsewhere in your
> molecule, you need to make sure that the nonbonded scaling factors for that
> parameter is "correct".
>
> --
> Jason M. Swails
> BioMaPS,
> Rutgers University
> Postdoctoral Researcher
> _______________________________________________
> 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
Received on Wed Oct 15 2014 - 20:00:03 PDT
