From: Jason Swails <>
Date: Fri, 29 Jul 2016 15:56:39 -0400

On Fri, Jul 29, 2016 at 3:27 PM, Thakur, Abhishek <> wrote:

> Hi
> I need some suggestion for MMPBSA calculation.
> First I have done MD for 10ns and then have carried out aMD for 300ns.
> Now I am planning to do MMPBSA .
> Now the place where I am struck is that in my MMPBSA calculation should I
> use the .crd files generated after 10ns of MD or from 300ns of aMD.

​You can't use the aMD trajectory directly. MMPBSA relies on the fact that
the conformations are sampled according to a Boltzmann distribution in
order to get properties that (in theory) correspond to thermodynamic
averages. If your conformational ensemble is *not* Boltzmann-weighted,
then this assumption fails and the resulting MM/PBSA energies will be

If you want to use your aMD trajectories for MM/PBSA, you will need to
reweight the energy time series the same way you reweight other time series
data coming from aMD.

Alternatively, you can use your 10 ns regular MD simulation (since it
*should* be Boltzmann-weighted), but if you needed aMD to sufficiently
sample your phase space, it's unclear how reliable a mere 10 ns *before*
employing enhanced sampling will be (since it will be strongly biased by
the free energy well it remains trapped in).

> Because I don't know the maths behind MMPBSA development, is it developed
> to take into account .crd files from aMD?

​The underlying theory is quite straightforward, and based on a fairly
simple thermodynamic cycle. As the manual suggests, I recommend running
MM/PBSA calculations "by hand" for at least a couple snapshots to ensure
you have a solid understanding of the method.

> ​​
> If not then will 10ns MD .crd will be sufficient to go for MMPBSA
> calculations?

​Depends. Maybe. Maybe not.


Jason M. Swails
AMBER mailing list
Received on Fri Jul 29 2016 - 13:00:04 PDT
Custom Search