[gmx-users] regarding the cosine content analysis

Mon Feb 25 13:32:22 CET 2013

Hi Thomas,

Okay :) First of all, do be careful that the first stage of
relaxation, giving that nice principal component profile, is not
exclusive. If you take it off, by leaving out the first part of the
trajectory, you may well find that there is another component which is
still relaxing, which maybe could only start after the first bit
finished (think of rotation and hinging; maybe have to hinge a bit
before the (other) rotation can relax). So checking cosine content
from the start up to the point that you have a low correlation and
then claiming that the system is relaxed after that is dangerous.

What that relaxation and equilibrium is? Let's started by introducing
the initial state T, as in T(ensed). That state can be a crystal
structure, in which the packing forces cause it to adopt a structure
that has a low probability under the conditions you're doing your
simulations at. It could also be the result from (homology) modeling,
where you can imagine a set of forces having acted to yield that
structure. Of course the same set of forces is still present, but
aimed outward from the protein if you change the environment. The
system is sort of wound up as a spring, and that potential energy
needs to dissipate before you can argue that the system samples
conformations that reflect the underlying, unperturbed energy
landscape. That's what the relaxation is: losing that initial strain
and get onto the energy landscape proper.
Equilibrium is of course a different matter still, where the actual
conformations and their relative probabilities come into play.

Cheers,

Tsjerk

On Mon, Feb 25, 2013 at 12:41 PM, Thomas Evangelidis <tevang3 at gmail.com> wrote:
> On 25 February 2013 12:14, Tsjerk Wassenaar <tsjerkw at gmail.com> wrote:
>
>> Hi Thomas,
>>
>> As I've explained previously, the cosine content does not allow such
>> inferences. Besides, taking the relaxation from the start into account
>> in PCA is pretty nonsensical, unless you aim to characterize that
>> relaxation in the first place. Looking at the cosine content to infer
>>
>
> Yes, that's what I meant, to identity those ns where the protein does
> unidirectional motion before it starts exploring the energy landscape, and
> exclude them from the final analysis (whatever this is). At least this is
> what I've seen doing in publications and this is what I have understood
> from our previous conversations. I think your objection is about the way I
> use the term "equilibration" to refer to what you call "relaxation" (I
> presume it is not the same as energy minimization). Perhaps it would be
> helpful to have your definition of the terms "equilibration" and
> "relaxation.
>
>
>> equilibration from that is blatantly fooling yourself.
>> If you feel you must use cosine content to support any claim on
>> equilibration, then it is a much better approach to start from the end
>> of the simulation and check that the stretch of the trajectory you
>> take does not yield high cosine contents, in which case you have some
>> reason to argue that that part of the simulation is sampled in a local
>> equilibrium.
>>
>> Cheers,
>>
>> Tsjerk
>>
>>
>> On Mon, Feb 25, 2013 at 10:56 AM, Thomas Evangelidis <tevang3 at gmail.com>
>> wrote:
>> > You don't do it the right way. You must start the analysis from the
>> > beginning not from the end of your trajectory. I.e.
>> >
>> > 0-20ns
>> > 0-30ns
>> > 0-40ns
>> > ...
>> > 0-100ns
>> >
>> > Until the cosine content of the first 3 principal components that account
>> > for most of the variance in the atomic fluctuation have been dropped at
>> > least once below 0.5. This is the point where theoretically the system
>> has
>> > equilibrated enough.
>> >
>> > Thomas
>> >
>> >
>> > On 22 February 2013 13:43, Ahmet yıldırım <ahmedo047 at gmail.com> wrote:
>> >
>> >> Dear users,
>> >>
>> >> I performed MD simulation of 400 ns of a structure. I used the cosine
>> >> content to check whether the simulation is not converged. I used last
>> 100
>> >> and 50 ns of trajectory to the analysis, respectively. The results were
>> >> very similar to each other.The cosine contents of the first ten
>> principal
>> >> components are as follows. The cosine contents of the principal
>> components
>> >> are very small but one. Why is the second cosine content differs from
>> the
>> >> others? What could be the reason for this? And do you think simulation
>> has
>> >> reached convergence?
>> >>
>> >> The cosine contents of last 50 ns:
>> >>  1 0.00685769
>> >>  2 0.137028
>> >>  3 0.00139929
>> >>  4 0.00903137
>> >>  5 0.0180072
>> >>  6 0.0128686
>> >>  7 0.00154502
>> >>  8 9.71793e-05
>> >>  9 0.00485945
>> >>  10 0.00202377
>> >>
>> >> Thanks in advance
>> >> --
>> >> Ahmet Yıldırım
>> >> --
>> >> gmx-users mailing list    gmx-users at gromacs.org
>> >> http://lists.gromacs.org/mailman/listinfo/gmx-users
>> >> * Please search the archive at
>> >> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>> >> * Please don't post (un)subscribe requests to the list. Use the
>> >> www interface or send it to gmx-users-request at gromacs.org.
>> >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>> >>
>> >
>> >
>> >
>> > --
>> >
>> > ======================================================================
>> >
>> > Thomas Evangelidis
>> >
>> > PhD student
>> > University of Athens
>> > Faculty of Pharmacy
>> > Department of Pharmaceutical Chemistry
>> > Panepistimioupoli-Zografou
>> > 157 71 Athens
>> > GREECE
>> >
>> > email: tevang at pharm.uoa.gr
>> >
>> >           tevang3 at gmail.com
>> >
>> >
>> > website: https://sites.google.com/site/thomasevangelidishomepage/
>> > --
>> > gmx-users mailing list    gmx-users at gromacs.org
>> > http://lists.gromacs.org/mailman/listinfo/gmx-users
>> > * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>> > * Please don't post (un)subscribe requests to the list. Use the
>> > www interface or send it to gmx-users-request at gromacs.org.
>> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>>
>>
>> --
>> Tsjerk A. Wassenaar, Ph.D.
>>
>> post-doctoral researcher
>> Biocomputing Group
>> Department of Biological Sciences
>> 2500 University Drive NW
>> Calgary, AB T2N 1N4
>> Canada
>> --
>> gmx-users mailing list    gmx-users at gromacs.org
>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>> * Please search the archive at
>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>> * Please don't post (un)subscribe requests to the list. Use the
>> www interface or send it to gmx-users-request at gromacs.org.
>> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>
>
>
> --
>
> ======================================================================
>
> Thomas Evangelidis
>
> PhD student
> University of Athens
> Faculty of Pharmacy
> Department of Pharmaceutical Chemistry
> Panepistimioupoli-Zografou
> 157 71 Athens
> GREECE
>
> email: tevang at pharm.uoa.gr
>
>           tevang3 at gmail.com
>
>
> website: https://sites.google.com/site/thomasevangelidishomepage/
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

-- 
Tsjerk A. Wassenaar, Ph.D.

post-doctoral researcher
Biocomputing Group
Department of Biological Sciences
2500 University Drive NW
Calgary, AB T2N 1N4
Canada