Hi
We've started to abandon the autocorrelation approach, just because
the problems with convergence. You have to look carefully on each
residue to see if they have really converged or not. Therefore, we
recommend to use iRED instead. On our test cases we haven't observed
any large overall difference between iRED and
autocorrelation-approach, but the latter requires more hands-on job.
/ Samuel
2010/5/1 Paul S. Nerenberg <psn.berkeley.edu>:
> Hi,
>
> Your e-mail is unclear in terms of what you mean by "negative value of the
> auto-correlation function". It is not unexpected that the auto-correlation
> function becomes negative at some correlation time (tau), although it should
> usually be fluctuating around zero when this happens (i.e., not reaching a
> large negative value). Moreover, if you were to average your C(tau)
> calculation over several trajectories, then you should see that these
> fluctuations about zero generally become smaller (as this "noise" is
> averaged out).
>
> So long as your simulations are adequate in length with regards to the max.
> correlation time and you are reasonably certain that your system can be
> analyzed with the model-free approach (which is nothing more than a *model*
> that makes assumptions about the dynamics of your molecule), then you should
> not have a problem fitting your auto-correlation functions to that
> exponential form.
>
> Best,
>
> Paul
>
> P.S. The max. correlation time used in your analysis (and correspondingly
> your simulation length) should always be chosen with regard to the
> characteristic relaxation times of your molecule, as is explicitly
> demonstrated in the article that Carlos referenced. I'm sure that the more
> NMR-minded AMBER users out there probably have some other pointers as
> well...
>
>
> On Apr 30, 2010, at 2:39 PM, luzhenw1.msu.edu wrote:
>
>> Thanks for the information. I read the Bruschweiler paper for the IRED
>> method and think it is OK to not use IRED approach. The question still
>> remains.
>>
>> Quoting Carlos Simmerling <carlos.simmerling.gmail.com>:
>>
>>> this is a useful article for such calculations
>>> http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2426661/
>>>
>>>
>>> On Fri, Apr 30, 2010 at 4:43 PM, <luzhenw1.msu.edu> wrote:
>>>
>>>> According to model-free formalism, the internal correlation function of
>>>> NH
>>>> vector, CI(t)=S^2+(1-S^2)e^(-t/tau)>=0. I tried to calculate the order
>>>> parameter from MD calculation. However, I got negative value of the
>>>> auto-correlation function for some residues. Is the defination of
>>>> autocorrelation different between NMR and MD? Thanks.
>>>> _______________________________________________
>>>> 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 Fri Apr 30 2010 - 23:00:03 PDT
