it's not quite clear what you're asking. I think that following Dan's
advice is good. make sure multiple simulations give you the same
distributions for the property you are asking about. meaning don't look at
average energy, or at base pairing in the stem, if you want to know if the
loop structure has converged. also, I strong suggest starting the
simulations in different conformations. short runs from the same structure
may all be stuck in the same place. in that case, consistent results are
not really meaningful. you need a reliable measure of precision. that will
be separate from the potential accuracy issues that you mentioned about
force field development. look at recent work from the Cheatham lab, among
others.
On Mon, Sep 1, 2014 at 1:15 PM, Asmita Gupta <asmita4des.gmail.com> wrote:
> This may sound stupid...but as far as i have read, most of the force field
> parameterization and torsion angle correction involves extensive testing on
> DNA/RNA duplexes. Testing on loops have been limited to tetraloops
> (UUCG/GNRA) which are very short and less prone to structural fluctuations
> (extensive base stacking and hydrogen bonding). If the loop length is long
> ( 7-8 nucleotides in my case) and all i am interested is in looking for
> stability of hydrogen bonds between these loops and duplex RNA stems, then
> how long is long enough? will 300-600ns give me a good estimate ? (provided
> the fact that i don't have access to a GPU machine, this long simulation
> will take ages to finish)
>
> Thanks for the replies...
>
>
> On Mon, Sep 1, 2014 at 9:20 PM, Carlos Simmerling <
> carlos.simmerling.gmail.com> wrote:
>
> > not really. imagine that you instead carried out 600,000 simulations of
> 1ps
> > each (still 600ns total). none will go far from the initial structure and
> > you would not expect to see motions that occur in 600ns. the issue is
> > somewhat complex, and depends a lot on whether you are looking at
> something
> > that obeys first orders kinetics and so on. the folding.home people have
> > done a lot of work in this area, since they tend to have many relatively
> > short simulations. Also, even if it is first order kinetics, the rate
> > constant will determine a probability of seeing an event in a certain
> > timescale. our simulations do not normally have the number of independent
> > events needed for them to show statistical behavior that might match up
> to
> > an experiment on a particular timescale. so, as Dave Case mentioned, if
> the
> > "events" that you see occur on a similar timescale to the length of the
> > simulation, you an't say much specific about whether another simulation
> > would or would not see the same behavior, or if it is significant that
> > another simulation does not.
> >
> > In your specific case, you should probably look into what others have
> seen
> > in their simulations of loops, since a fair number of studies have been
> > published in that area. I agree with DAC, 200 or even 600ns is not very
> > long. Much depends on the initial structure and what type of dynamics you
> > are trying to study.
> >
> >
> > On Mon, Sep 1, 2014 at 9:38 AM, Asmita Gupta <asmita4des.gmail.com>
> wrote:
> >
> > > But, then i have 3 different runs for the same system, each 200ns. Is
> it
> > > incorrect to say that the system has experienced conformational
> sampling
> > > for a total of 600ns, which might be long enough for loops?
> > >
> > > Asmita
> > >
> > >
> > > On Mon, Sep 1, 2014 at 5:54 PM, David A Case <case.biomaps.rutgers.edu
> >
> > > wrote:
> > >
> > > > On Mon, Sep 01, 2014, Asmita Gupta wrote:
> > > > >
> > > > > In 2 out of 3 trajectories, the structure is experiencing a loop
> > > residue
> > > > > backbone flipping (zeta, chi and delta), after which it remains
> > stable.
> > > > > However, with parmbsc0, this residue flipping is not observed at
> all,
> > > > > neither with YIL chi correction.
> > > >
> > > > One useful analysis is to see how long it took for the flip to take
> > place
> > > > in
> > > > the two trajectories where it did happen. If these are significant
> > > > fractions
> > > > of the 200 ns length, then it is plausibly just a matter of chance
> that
> > > you
> > > > did not see such behavior in the third run. You could extend the
> third
> > > run
> > > > longer to see what happens.
> > > >
> > > > >
> > > > > is 200ns enough for a good conformational sampling?
> > > >
> > > > This is not sufficient in general to sample RNA loop configurations.
> > > >
> > > > ....dac
> > > >
> > > >
> > > > _______________________________________________
> > > > 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
> > >
> > _______________________________________________
> > 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
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Sep 01 2014 - 12:00:03 PDT
