Re: [AMBER] capturing free energy landscape

From: Adrian Roitberg <>
Date: Tue, 10 Dec 2013 16:26:58 -0500

BTW: careful with PCAs...

First, make sure that the PCAs you choose represent a substantial amount
of the variance. Too many papers show 1 or 2 PCAS and the sum of them is
only 20% of the variance. Granted, it is the top 20%, but even then, it
is a small part of the story.

Second, PCAS are notoriously hard to converge. Do this: take your 75 ns
MD, split in into 2 halves. Redo PCA for each half separately and see if
the first PCA vector for one half is similar to the first PCA vector for
the second one. If not, then your PCA vecotr for the free energy is very
arbitrary and does not tell you too much about your system.


On 12/10/13 3:59 PM, Thomas Evangelidis wrote:
> Carma does what Jason described automatically, namely you can plot the Free
> Energy along two or three Principal Components. All you need is to convert
> your .prmtop/.nc to .psf/.dcd with cpptraj from AmberTools. Check it out:
> HTH,
> Thomas
> On 10 December 2013 22:45, Jason Swails <> wrote:
>> On Tue, Dec 10, 2013 at 5:23 AM, anu chandra <> wrote:
>>> Dear Amber users,
>>> I am working with all-atom MD simulations of protein-ligand system. I
>> have
>>> carried out PCA analysis after 75ns of simulation. By looking at the
>>> collective motions along the first principle component, I could able to
>> see
>>> that the protein is visiting two different conformations.Now, I am
>> looking
>>> for a method to capture the free energy landscape of these conformational
>>> changes. If I am right, usual methodologies like meta-dynamics. AMD etc.
>>> are usually used to overcome the high energy barrier during the
>>> conformational changes. Since I could capture the conformational changes
>> in
>>> my protein during the classical MD itself, it seems like the energy
>> barrier
>>> is too low between these two different conformations. How can I capture
>>> free energy change during these conformational changes? Which method will
>>> be helpful for me to do this calculation?
>> If you have sufficient sampling without using any type of biasing
>> potential, you can just calculate free energies directly from the
>> simulation.
>> This is typically done using a histogram along whatever reaction coordinate
>> you choose to define and taking the negative log of the histogram
>> population (divided by the population of the most populated bin if you want
>> to set the minimum energy to 0). Multiplied by -kT, of course...
>> Good luck,
>> Jason
>> --
>> Jason M. Swails
>> BioMaPS,
>> Rutgers University
>> Postdoctoral Researcher
>> _______________________________________________
>> AMBER mailing list

                             Dr. Adrian E. Roitberg
Colonel Allan R. and Margaret G. Crow Term Professor.
Quantum Theory Project, Department of Chemistry
University of Florida
AMBER mailing list
Received on Tue Dec 10 2013 - 13:30:03 PST
Custom Search