[AMBER] TI calculations and autocorrelation time

From: Kepa K. Burusco <kekoburgo.yahoo.es>
Date: Thu, 27 Jun 2013 15:56:48 +0100 (BST)

[27-VI-2013] Hi everybody, I am running TI calculations with amber12. I am mainly following Steinbrecher's paper J. Comput. Chem. (2011). 32(15). 3253-3264 as a reference to apply this technique. In order to correctly estimate the errors in the individual lambda states I am following equations 10 and 11. I was interested in determining the autocorrelation time to correct/compensate the effect of the memory of the system (MD data sets are partially correlated) when calculating the errors, so I have written a program in Java to post-process the DV/DL lines in the output file. The autocorrelation time is calculated according to equations in: Mol. Phys. (1986). 57(1). 89-95 and J. Chem. Theory. Comput. (2007). 3(1). 26-41 And the Standard Error is calculated according to equations in: J. Chem. Phys. (2007). 127. 214108 and J. Comput. Chem. (2011). 32(15). 3253-3264 My questions are: 1) In the 2011 paper it is said that a typicall value for the autocorrelation time is something in the range of 1 ps but in my calculations (about 2 ns length) my autocorrelation times are really larger than that value. I am using samplings of 1 conformation/ps and 100 conformations/ps and in both cases the autocorelation times range roughly from 10ps to 200ps (see attached text file). I am using timespets of 1 fs and langevin thermostat (ntt=3) with gamma_ln=2. Does it make any sense? 2) Do you use your own "in-home" programs to calculate the autocorrelation time or is there any recommended sorftware to do it? I tested my own Java program with a pure discrete periodic signal [y = sin(2*pi*x/100)] and I obtained an autocorrelation time of about 25, which makes sense since the data set is orthogonal at that value of the time shift. However, still I would like to check that my software is working correctly by contrasting the results with other program/script. Any help is welcome. Thank you very much in advance. Kepa K. *********************************************************

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber

Received on Thu Jun 27 2013 - 08:00:02 PDT
Custom Search