On Tue, Aug 27, 2013 at 7:57 PM, <psu4.uic.edu> wrote:
> Dear Amber,
>
> We have a question about calculating correlation time of mmP(G)BSA dG
> values. After consulting the amberTools 13 manual, old mailing
> list<http://archive.ambermd.org/200712/0070.html>and the manuscripts
> using the statistical inefficiency method (J. Chem.
> Phys. 120, 2618 (2004); doi: 10.1063/1.1638996 & J Comput Chem 31:
> 837–846, 2010 ), we assume the following way is correct to calculate
> correlation time of dG in mmP(G)BSA?
>
> 1. Run MD simulation and save the trajectory in every picosecond (ps).
>
> 2. Calculate mmP(G)BSA dGbind using every ps snapshot from the MD
> trajectory. Save the .csv file during mmP(G)BSA (Activate -eo flag).
>
> 3. Write a script to calculate the dGbind correlation time using dGbind
> of each snapshot from the .csv file by the statistical inefficiency method.
>
> * statistical inefficiency method: correlation time= (block length) *
> (the variance of the block average of X) / (the variance of the
> distribution (X))
>
This way will work, but there is an easier method if you plan to use Python
to analyze your data (with, e.g., numpy, scipy, pandas, etc.). See pages
376--377 of the AmberTools 13 manual where it describes using the Python
API giving an example where the autocorrelation function is printed.
The binding energy can be computed pretty easily by just subtracting the
receptor and ligand TOTALs from the complex TOTAL (numpy arrays follow
standard vector add/subtract rules). This approach prevents you from
having to write a separate script to parse the CSV file. The example in
the manual demonstrates how you would calculate the normalized
autocorrelation function (which can also be used to estimate the
statistical inefficiency in which the function becomes zero within some
range of tolerance) and plots the resulting autocorrelation function with
matplotlib, but you can adapt it for your own purposes.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 28 2013 - 05:30:09 PDT
