AMBER Archive (2008)

Subject: Re: AMBER: TI calculations in Amber10

From: Thomas Steinbrecher (
Date: Wed Jun 25 2008 - 11:15:47 CDT

Hi Samuel,

from what you describe it doesnt look like something in your calculation
is 'wrong' (i.e. using Amber in a way it is not supposed to work), the
steps you describe seem to be appropriate. Please note that high dVdl
values in the middle step are not unusual, since you are changing not only
the softcore atoms (which probably give rise to large dVdl fluctuation on
their own) but also every charge on a non-softcore atom. They should
however more or less cancel out when you subtract the water and protein

One little thing, it looks like in your input file that you have a 0
(as in zero) in crgmask where you should have an O (as in oxygen).

Some things that might have gone wrong (i.e. why is your result not close
to the experimental one) could be that the two ligands dont share the same
binding mode or that the protein would need to adjust to the new ligand by
conformational changes that dont happen on the timescale of your
simulation. Do you get a much to unfavorable binding energy for the ligand
you change into? That could indicate such a sampling problem. Take a look
at your trajectories if there are obvious clashes for the new ligand. Also
test if your simulations are really converged, say run another .5 ns at
each lambda and check if you get almost the same deltaG result from those
new simulations. Also take a look at your dVdl curves, are they really
smooth or should you add some more l-windows somewhere.



Dr. Thomas Steinbrecher
The Scripps Research Institute
10550 N. Torrey Pines Rd.
San Diego CA 92037, USA

On Wed, 25 Jun 2008, Samuel Genheden (a03samge) wrote:

> Hello, Amber users
> javascript:SetCmd(cmdSend);
> I'm trying to calculate relative binding free energies of two drug
molecules using the TI facilities in Amber10. The perturbation is an
O-atom to a NH-group. I've followed the tutorial "small molecule binding
to T4-lysozyme" and therefore breaked up the perturbation in 3 steps: 1)
remove charge of O-atom, 2) transformation of neutral O-atom to neutral
NH-group with soft-core potential and 3) addition of charge to NH-group.
I have used 9 lambda values from 0.1 to 0.9. The experimental difference
in binding energy is very small, in the order of 1 kcal/mole. The 1:st
and the 3:rd steps seemed to work okey since the deltaG are on the order
of 0.1 kcal/mole. However the problem is in the 2:nd step where I get
unreasonable high dv/dl values, both in the protein and in the water
phase. Here the dv/dl are about 15-30 kcal/mole and deltaG evaluates to
approx. -20kcal/mol. Here is an example of the input file for the
production run:
> Production
> &cntrl
> irest=1,ntx=5,
> nstlim=1000000,dt=0.002,
> temp0=300.0,ntt=3,gamma_ln=2.0,
> ntc=2,ntf=1,
> cut=8.0,
> ntpr=5000,ntwx=5000,ntwv=0,ntwe=5000,iwrap=1,
> ntb=1,
> ipol=0,igb=0,
> scnb=2.0,scee=1.2,
> ntr=0,
> icfe=1, clambda = 0.5,
> ifsc=1,
> crgmask=':CBB_at_032',
> scmask=':CBB_at_O32',
> &end
> I hope someone can shed some light on my problem, and maybe point to possible sources of error.
> Best regards,
> Samuel
The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)