[gmx-users] Implicit Solvent Gromacs 4.5.5, LINC errors in pulling simulations

[mhabibi at phas.ubc.ca]

Hi,

I am running a pulling simulation with implicit solvent using Gromacs
4.5.5, single precision on 8 CPU, with OPLS-AA FFT.

 The system is equilibrated for 10 ns, with pbc=0 and infinite cut-off
(.mdp file is posted). I do not get any LINC errors for the
equilibration, however when I start pulling the protein with infinite
cut-off, after 10-15 nm extension (this is about 1500 ps with pulling
rate=0.01 nm/ps), the system crashes due to the LINC errors. For
different runs, the LINC errors are for different atoms.

I have tried pulling rates =0, 0.001, 0.01, and 0.1 nm/ps, but even for
pulling rate=0, I always get the LINC errors with infinite cut-off in
implicit solvent.

When I use a finite cut-off distance (4, 6, 8, 10, and 20 nm), there is no
LINC error and I can unfold the protein to the full extension.  If I run
the simulation with cut-off of 40  nm, I get the same errors as infinite
cut-off. The size of the folded protein is less than 4 nm, the unfolded
protein is about 35 nm. I ran a one step simulation for a folded protein,
and the GB and energy terms are the same for cut-off distances of
6,8,10,20,30,40 and Infinit.

There is no LINC error in vacuum dielectric=80 with infinite cut-off.

My question is if there is something wrong in my .mdp file? or this is an
issue with GB calculation?
Is there any way to avoid LINC errors in the pulling simulation with
implicit solvent and infinite cut-off?

For version Gromacs 5.0.4, I have some issues with running the simulations
in parallel, so version 4.5.5 is the only available option.

Thank You,
Mona



; Run parameters
integrator  = sd
ld_seed     =-1
dt          = 0.001
tinit       = 0
nsteps      = 5000000
nstcomm     = 10
; Output parameters
nstxout     = 1000
nstvout     = 1000
nstfout     = 1000
nstxtcout   = 1000
nstenergy   = 1000

; Bond parameters
constraint_algorithm = lincs
constraints     = h-bonds     ; H bonds constrained
lincs_iter        = 1                 ; accuracy of LINCS
lincs_order      = 4               ; also related to accuracy


continuation      = yes      ; continuing from NVT

comm_mode = ANGULAR

cutoff_scheme = group
pbc                 = no      ; Periodic boundary conditions disabled
coulombtype    = cut-off ; Calculate coulomb interactions using cutoff
rcoulomb         = 0.0     ; Coulomb cutoff of infinity
vdw_type        = cut-off ; Calculate van der Waals interactions using cutoff
rvdw               = 0.0     ; Van der Waals cutoff of infinity
rlist                 = 0.0     ; Neighbor list cutoff
nstlist             = 0.0       ; Do not update neighbor list


tau_t       = 2
ref_t       = 300


implicit_solvent     =   GBSA
gb_epsilon_solvent   =   80
gb_saltconc          =   0.10
gb_algorithm         =   OBC
rgbradii             =   0.0
sa_surface_tension   =   2.25936


; Pull code
pull            = umbrella
pull_geometry   = distance  ; simple distance increase
pull_dim        = N N Y
pull_start      = yes       ; define initial COM distance > 0
pull_ngroups    = 1
pull_group0     = Freeze
pull_group1     = Pull
pull_rate1      = 0.001      ;  nm per ps
pull_k1         = 1000      ; kJ mol^-1 nm^-2
pull_nstxout    = 1000
pull_nstfout    = 1000