[AMBER] TI electrostatic free energy

From: Franz Waibl <Franz.Waibl.uibk.ac.at>
Date: Mon, 30 Nov 2020 15:34:44 +0100

Dear Amber community,

I am trying to compute the free energy of hydration of a set of small
molecules at different salt concentrations, using TI. I want to use a
2-step transform (i.e., first remove the charges, then the Van der Waals

However, I have problems figuring out how to set up the electrostatic
step in pmemd (CUDA-version, using Amber 20). Currently, my input file
looks like this:

production 10 ns npt run - Berendsen barostat, TI, remove charges
     ntx=1, irest=0,
     ntb=2, iwrap=1,
     ntp=1, barostat=2, pres0=1.0,taup=1.0,
     ntc=2, ntf=1,
     ntt=3, tempi=300.0, temp0=300.0, gamma_ln=2,
     nstlim=5000000, dt=0.002,
     ntwr=50000, ntwx=5000,
     icfe=1, klambda=1, ifsc=0, timask1=':1', timask2=':2', crgmask=':2',
     ifmbar=1, mbar_states=11, tishake=0,

(Of course, I have equivalent input files for the other lambda values).
Of this run, I am discarding the first 40% (4ns) as equilibration. The
1st and second residue are 2 copies of my molecule (the topology was
created using tleap by "combine { solute solute solvent }".)

The problem is that I am getting very large free energies (e.g, 130
kcal/mol for caffeine, favoring lambda=0). This occurs with and without
the salt, so I think the problem is not related to the ions.
Furthermore, I don't think that convergence is the problem, because I
tried using 10x longer simulations for a few systems, and the results
are very similar.

The VdW contribution is rather close to zero, and therefore doesn't
compensate the high values. I assume that this is realistic, and that
the problem is related to the electrostatics part. (is this correct?)

I would be very thankful for any suggestions on what might be wrong here...

Best regards,

Franz Waibl

AMBER mailing list
Received on Mon Nov 30 2020 - 07:00:03 PST
Custom Search