[gmx-users] Interpretation of g_hbond results

[David van der Spoel]

Justin A. Lemkul wrote:
> 
> 
> David van der Spoel wrote:
>> Justin A. Lemkul wrote:
>>>
>>> Hi,
>>>
>>> I have a quick question about interpreting the output from g_hbond.  
>>> I am using the -ac option to calculate H-bond lifetimes between my 
>>> protein and a series of different, bound ligands.  I read the 
>>> associated paper, but I would like to confirm the interpretation of 
>>> the results before I continue much further.  I obtained the following:
>>>
>>> --------------------------------------------------
>>> Type      Rate (1/ps) Time (ps)  DG (kJ/mol)  Chi^2
>>> Forward         0.003    309.475      19.590  0.000294756
>>> Backward        0.007    136.034      17.471
>>> One-way         0.002    471.990      20.678
>>> Integral        0.001   1534.894      23.717
>>> Relaxation      0.002    606.619      21.325
>>>
>>> The "forward" and "backward" rates and times make sense based on 
>>> equations 3 & 4 in the paper, but I could not find an explanation of 
>>> what "one-way" and "relaxation" are.  Is the "integral" value taken 
>>> from equation 2, and thus should be interpreted as the overall H-bond 
>>> lifetime?
>>>
>>> Thanks,
>>> Justin
>>>
>> One-way assumes k' (Backward) rate is zero, Relaxation is the time at 
>> which C(t) = 1/e
>> Integral is just the integral of C(t), which in case of exponential 
>> relation should be the same as the relaxation time (not very close as 
>> you see).
>>
> 
> Thanks, that makes sense.  According to the terminology in your paper, 
> then, if k is the rate constant for HB breaking ("forward"), then 1/k is 
> the lifetime of the hydrogen bond.  So that lifetime would be the time 
> (in ps) given in the "Forward" entry?  It seems simple, I just want to 
> be sure since there are a couple of old threads in the archive where 
> users claimed that "Backward" was the value to use.
> 
Forward is for breaking, so yes, that is the life time recommended by 
Luzar2000.

>> Two caveats: first, we recently found out that the numbers in my JPCB 
>> paper can not be reproduced in 4.0 (embarassingly this was found out 
>> during a tutorial that I was giving). We are looking into this. Second 
>> we are busy implementing a new algorithm due to Markovitch * Agmon 
>> (JCP 129 (2008) p. 084505) that has several advantages over the Luzar 
>> approach.
>>
> 
> My simulations were conducted under 4.0.2, should I not use these 
> results? Results from 3.3.x should be correct, since it was used in your 
> JPCB paper; but my installation of 3.3.3 gives a segmentation fault on 
> our cluster, and a bus error on my local machine.  Are there known 
> issues with g_hbond-3.3.3?

As said, we are looking into this. With the original trajectories from 
the paper the 4.0 code gives different values. I haven't tried with 3.3 
analysis tools yet. Digging back into my memory, the problem seems to 
have to do with the correlation function rather than the analysis, that 
is, you can do the Luzar analysis in xmgrace. based on the ACF that 
g_hboond spits out. If you do so you find that you get the same results 
as g_hbond prints, ergo the Luzar analysis is implemented OK, but the 
ACF is suspect.

> 
> Or should I just hold off and do analysis after a new version comes out 
> and these issues are solved or a new algorithm implemented?

This is up to you to decide...

> 
> Thanks again,
> Justin
> 
>> Cheers,
> 


-- 
David.
________________________________________________________________________
David van der Spoel, PhD, Professor of Biology
Dept. of Cell and Molecular Biology, Uppsala University.
Husargatan 3, Box 596,  	75124 Uppsala, Sweden
phone:	46 18 471 4205		fax: 46 18 511 755
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++