[gmx-users] fep: continuation of run after lambda = 1

Evelyne Deplazes 20400454 at student.uwa.edu.au
Fri May 14 05:15:51 CEST 2010


Hi

I am simulating a system consisting of a protein embedded in membrane,
solvated with water using the Martini forcefield. I have set up a free
energy perturbation simulation where a subset of the protein particles are
being converted from their original particle type (state A, lambda =0) to a
new particle type (state B, lambda =1). I have finished a 500 ns run with
the following fep options

free_energy = yes
init_lambda = 0.0
delta_lambda = 5.0e-8

the no. of steps of the simulation is

nsteps              = 20000000
dt                  = 0.025

hence lambda = 1 after 500ns. so in theory the confout.gro should contain
the structure with all the "fep" particles in state B. I now want to
continue the simulation with the new particle types without any fep options
(ie a simple md run).

I have tried the following to continue my simulation:

1. I used the same *itp file and a new .mdp file with

free_energy = yes
init_lambda = 1.0
delta_lambda = 0.0

gen_vel = 0

I grommp-ed the new .mdp file (using the same *top, *itp and *gro file as
for the original grompp) and continued the run using the state.cpt files
from the end of the 500ns run.

The run crashed with the following msg

----------------------------------------------------------------------------------------------------------------------------------------------
Step 20000022, time 500001 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms inf, max inf (between atoms 345 and 346)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
    345    346   90.0    0.2700      inf      0.2700
    345    347   90.0    0.2700      inf      0.2700
    346    347   90.0    0.2700      inf      0.2700
    354    355   90.0    0.2650      inf      0.2650
----------------------------------------------------------------------------------------------------------------------------------------------

Note: atoms 345, 346 and 347 are particles that have been changed to a new
type during the fep run.

2. I made a new .itp file where I changed all the "fep" particles to be of
the new type (ie state B) and removed all fep parameters in the .mdp file. I
"re-grompped" with the confout.gro from the fep run and the new  .itp and
.mdp files.  This should be like a new simulation without any "memory of its
fep past".

The run crashed with the following msg
----------------------------------------------------------------------------------------------------------------------------------------------
starting mdrun 'MSCL PROTEIN IN MEMBRANE, SOLVATED, WITH IONS'
20000000 steps, 500000.0 ps.
step 0
step 100, remaining runtime:     0 s
step 200, remaining runtime:     0 s

Step 264, time 6.6 (ps)  LINCS WARNING
relative constraint deviation after LINCS:
rms 4.088986, max 75.397156 (between atoms 455 and 456)
bonds that rotated more than 30 degrees:
 atom 1 atom 2  angle  previous, current, constraint length
    464    465   49.1    0.2700   0.2669      0.2700
    464    466   49.4    0.2700   0.2829      0.2700
    455    456   90.0    0.2650  20.2452      0.2650

----------------------------------------------------------------------------------------------------------------------------------------------

I also tried to simply minimize the confout.gro (ie the structure after the
fep run has finished) but get the following

Double precision normally gives you higher accuracy.
You might need to increase your constraint accuracy, or turn
off constraints alltogether (set constraints = none in mdp file)

writing lowest energy coordinates.

Steepest Descents converged to machine precision in 99 steps,
but did not reach the requested Fmax < 10.
Potential Energy  = -1.0321324e+06
Maximum force     =  4.4505771e+03 on atom 462
Norm of force     =  3.2656403e+01


I thought it might be that I am "re-grompping" and starting the simulation
again. So I set up a test run where there is a continuous simulation. Total
simulation time = 200ns and the fep options are set such that lambda reaches
1 after 100 ns. I was hoping the simulation would simply continue with the
new particle type for the remaining 100 ns. But that crashed as well with
similar LINCS warnings.

Has anyone else ever had a similar problem? I have already spent a fair bit
of time on trying to fix it and also talked to people from the Martini group
who are experiences gromacs user.

any ideas are appreciated

Thanks
-- 
Evelyne Deplazes

PhD student
Theoretical Chemistry group
University of Western Australia
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20100514/5db983e4/attachment.html>


More information about the gromacs.org_gmx-users mailing list