Re: [AMBER] Problem with TI minimization

From: David A Case via AMBER <>
Date: Fri, 2 Sep 2022 16:17:12 -0400

On Thu, Sep 01, 2022, Matthew Guberman-Pfeffer via AMBER wrote:

A few comments are given below. But just seeing three total energy numbers
doesn't provide enough information to help much.

>My minimization for a lambda window in TI is behaving oddly. The energy
>first decreases from 6.1363E+06 to 1.5983E+06 and then increases to
>1.6415E+06 over 10000 steps. Below is my input. timak and scmask define the
>enter residue being mutated. The troubling result is the same if I restrict
>the softcore region to only the appearing /disappearing atoms. What am I
>doing wrong?
> &cntrl
> imin = 1, ! minimization on
> ntmin = 2, ! Only the steepest descent method is used

    Running steepest descent only is almost aways a very bad idea. Consider
    sticking to the default.

> ncyc = 5000, ! minimization will be switched from steepest descent to conjugate gradient

   This is just wasting 5000 steps of minimization; stick with the default,
or at least a value two orders of magnitude smaller. (Of course, with
ntmin=2, the value of ncyc is ignored.)

> dx0 = 1.0D-7, !The initial step length. If the initial step length is too big then will give a huge energy

    This is just slowing you down as well; consider sticking to the default

> ntr = 1, restraint_wt = 100.00,
> restraintmask='!:WAT',

   This is very high restraint weight; are you sure that you don't want any
of the waters to move?
> icfe = 1, !free energy calculations
> ifsc = 1, !1 SC potentials are used

Have you minimized and equilibrated one the end states first?


AMBER mailing list
Received on Fri Sep 02 2022 - 13:30:03 PDT
Custom Search