[gmx-users] Re Warning- Correlations not long enough

[David van der Spoel]

Den 2018-09-11 kl. 17:54, skrev Apramita Chand:
> Dear Dr.Spoel,
>   As suggested, I plotted the correlation function in xmgrace and it looks
> fine. I used -fitfn option and also -smooth option ( which  If >= 0,
> smooths the tail of the ACF by fitting it to an exponential function: y =
> Aexp(-x/tau)
> 
> Surprisingly, for all these options (or without) the same warning shows up
> and the hbond lifetime obtained remains unchanged.
> 
> WARNING: Correlation function is probably not long enough
> because the standard deviation in the tail of C(t) > 0.001
> Tail value (average C(t) over second half of acf): 0.00340811 +/- 0.00108356
> 
> Back Off! I just backed up hbac_30_pt.xvg to ./#hbac_30_pt.xvg.3#
> Hydrogen bond thermodynamics at T = 300 K
> Fitting parameters chi^2 = 0.00596444
> 
> Should I go with the lifetime obtained?

The number is quite small, and if is this for one specific hydrogen bond 
it will not get much better.
> 
> Yours sincerely,
> Apramita
> 
> 
> reply to :
> Message: 2
> Date: Sat, 8 Sep 2018 08:32:43 +0200
> From: David van der Spoel <spoel at xray.bmc.uu.se>
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] Warning: correlations not long enough
> Message-ID: <eb489850-a054-2e2b-e062-f72e29b6feaa at xray.bmc.uu.se>
> Content-Type: text/plain; charset=windows-1252; format=flowed
> 
> Den 2018-09-07 kl. 17:33, skrev Apramita Chand:
>> Dear Dr. Spoel,
>> Initially I used a 100ns simulation with 0.002 time step.
>>    Then I tried a 10ns simulation with 0.0005 timestep and lower output
>> frequency.
>>    I used -ac option and obtained the forward lifetime.
>>
> That looks good, if you open the autocorrelation figure in xmgrace, does
> it look good? Can you fit an exponential curve to it with a time
> constant roughly  like 45 ps?
> 
> If you are analyzing a HBond that is present during the whole simulation
> the analysis will not give a meaningful result.
>>
>>
>> Regards,
>> Apramita
>>
>>
>>
>> Reply to :Message: 2
>> Date: Fri, 7 Sep 2018 08:14:48 +0200
>> From: David van der Spoel <spoel at xray.bmc.uu.se>
>> To: gmx-users at gromacs.org
>> Subject: Re: [gmx-users] Warning: correlations not long enough
>> Message-ID: <ad407060-789d-bc24-354d-cbf089b8a9c7 at xray.bmc.uu.se>
>> Content-Type: text/plain; charset=windows-1252; format=flowed
>>
>> Den 2018-09-06 kl. 17:13, skrev Apramita Chand:
>>> Dear All,
>>> In my simulations ,I have a Peptide in aqueous solution with a cosolvent.
>>>
>>> For more number of hydrogen bonds like peptide water, water-water, there
>>> are no warnings
>>>
>>> But for Peptide cosolvent, there are 315 hydrogen bonds and it does the
>> acf
>>> as 315/315...
>>>
>>> It shows warning: correlations not long enough
>>>
>>> My original time step was 0.002 and I changed it to 0.0005 and even
>> lowered
>>> the output frequency of coordinates
>>>
>> How long is the simulation? You do not have to store so often, may 50 fs
>> is enough, but you may need a longer simulation. If 45 ps is a realistic
>> correlation time you need at least some nanoseconds. And do not use the
>> -life option, as is described in my old JPCB 110 pp. 4393-4398 (2006).
>>
>>> Doing so gave me a 80GB output file!
>>>
>>> Still the problem persists..
>>>
>>> The lifetime that is output on the screen is 45 ps.
>>>
>>> Since it writes that the lower half of the correlation function was
> fitted
>>> and then this is calculated, can I quote this lifetime? Is it accurate?
>>>
>>>
>>>
>>> Please help
>>>
>>>
>>> Yours sincerely
>>> Apramita
>>>


-- 
David van der Spoel, Ph.D., Professor of Biology
Head of Department, Cell & Molecular Biology, Uppsala University.
Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
http://www.icm.uu.se