Re: How to deal with a long path in ABF simulation?

From: Wenyu Zhong (wenyuzhong_at_gmail.com)
Date: Thu Apr 27 2006 - 07:28:34 CDT

Dear Chris,

Thanks very much for your advice.

The free energy profile I am interested in is the translocation of a toxin
across a carbon nanotube. The whole reaction path is 44 A long, from
the mouth to the center of the CNT.

I have run sevreal simulations and different results were obtained.

First, a sing 10 ns run.
abf parameters:
coordinates Zcoord
dxi 0.2
xiMin -44.0
xiMax 0.0
forceconst 100.0
dSmooth 0.4

The peak of the free energy get up to 60 Kcal/mol.

Second , the path was splited into 22 contiguous, non-overlapping
parts.Other parameters were unaltered. Every single simulations was
0.5ns long. After combining all the results, a peak of 25 kcal/mol was
obtained.

To refine the results from the second simulations, I did the third
simulations. The bins sampling less than 10000, were simulated again in
shorter windows. These simulations haven't been completed till now. By
combining the results with previous results, I guess the peak would be about
10 kcal/mol.

Obviouslly, a single run is not enough, and sevreal runs may be not enough,
either. However, Is the way right?

The settings of ABF in the guide were not easy to understand for me. To make
use of ABF, I have to learn, guess, and practise again and again. How to do
"a single, large window, using the biases inferred from the multistage free
energy calculation"? Is it using the setting of "moveBoundary"?

I think a good strategy is changing the xiMin and xiMax in every bin. Is
that the "moveBoundary" meaning?

Wating for your response.

Regards

Wenyu

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:41:59 CST