convergence criterion of free energy perturbation (FEP) simulations for pKa calculation

From: Sadegh Faramarzi Ganjabad (
Date: Thu Oct 05 2017 - 13:36:07 CDT


I am running a test FEP simulation to calculate free energy of the
ionization of an amino acid inside a membrane protein. Here is the FEP part
of my namd config.

source ../tools/fep.tcl

alch on
alchType FEP
alchFile all-ion-beta.pdb
alchCol B
alchOutFile forward-shift-pr.fepout
alchOutFreq 10
alchBondDecouple on
alchBondLambdaEnd 1.0

alchVdwLambdaEnd 1.0
alchElecLambdaStart 0.5 #soft potential
alchVdWShiftCoeff 6.0
#alchDecouple off

alchEquilSteps 100
set numSteps 500

runFEPmin 0.0 1.0 0.0625 10000 1000 310

Then I used ParseFEP of VMD. The difference between DeltaG of the forward
and backward transformation is about 5 kcal/mol for lambda=0 to 0.5, and it
become almost 0 for lambda = 0.6 to 1. I should probably increase
simulation time for each lambda from 10,000 to say 10,000,000 until the
plots of forward and backward lie on top of each other. My question is that
is this the only criterion for convergence of FEP calculation? How can I
make sure my simulation is long enough?

Also, if I use thermodynamic integration method for free energy
calculations, what would be convergence criteria for that?


This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:42 CST