[gmx-users] mdp file help

[Muhammad Ayaz Anwar]

Hi guys, I'm new to Gromacs and seeking some input on my .mdp file for the production run. I want to perform simulation to check protein stability over time and the after that the stable protein will be used for protein-protein docking/interactions. I'm using cubic box (with 1nm pbc) with water and Cl ions to neutralize on a GPU accelerated system. my mdp file is as: title	 = Protein in water ; Run parameters integrator	= md	 ; leap-frog integrator nsteps	= 2000 ; 2 * 500000 = 1000 ps, 1 ns dt	 = 0.002	 ; 2 fs cutoff-scheme     = Verlet     ; for GPU acceleration verlet-buffer-drift = -1	        ; now use nstlist ; Output control nstxout	 = 1000	 ; save coordinates every 2 ps nstvout	 = 1000	 ; save velocities every 2 ps nstxtcout	        = 1000	 ; xtc compressed trajectory output every 2 ps nstenergy	        = 1000	 ; save energies every 2 ps nstlog	 = 1000	 ; update log file every 2 ps ; Bond parameters continuation	= yes	 ; Restarting after NPT constraint_algorithm = lincs	; holonomic constraints constraints	            = all-bonds	; all bonds (even heavy atom-H bonds) constrained lincs_iter	            = 1	 ; accuracy of LINCS lincs_order	            = 4	 ; also related to accuracy ; Neighborsearching ns_type	    = grid	 ; search neighboring grid cells nstlist	    = 30	 ; 10 fs rlist	            = 0.6	 ; short-range neighborlist cutoff (in nm) rcoulomb     = 0.6	 ; short-range electrostatic cutoff (in nm) rvdw	            = 0.6	 ; short-range van der Waals cutoff (in nm) ; Electrostatics coulombtype	  = PME	 ; Particle Mesh Ewald for long-range electrostatics pme_order	          = 4	 ; cubic interpolation fourierspacing	 = 0.12	 ; grid spacing for FFT ; Temperature coupling is on tcoupl	 = V-rescale	; modified Berendsen thermostat tc-grps	 = Protein Non-Protein	; two coupling groups - more accurate tau_t	 = 0.1	0.1	; time constant, in ps ref_t	 = 300 300	; reference temperature, one for each group, in K ; Pressure coupling is on pcoupl	 = Parrinello-Rahman	; Pressure coupling on in NPT pcoupltype	= isotropic	; uniform scaling of box vectors tau_p	 = 2.0	 ; time constant, in ps ref_p	 = 1.0	 ; reference pressure, in bar compressibility = 4.5e-5	; isothermal compressibility of water, bar^-1 ; Periodic boundary conditions pbc	 = xyz	 ; 3-D PBC ; Dispersion correction DispCorr	= EnerPres	; account for cut-off vdW scheme ; Velocity generation gen_vel	 = no	 ; Velocity generation is off By performing a test run with ff99SB force field, I got an efficiency as follow: Reading file test11.tpr, VERSION 4.6.1 (single precision) Using 1 MPI thread Using 4 OpenMP threads 1 GPU detected:   #0: NVIDIA GeForce GT 630, compute cap.: 3.0, ECC:  no, stat: compatible 1 GPU auto-selected for this run: #0 starting mdrun 'Protein in water' 2000 steps,      4.0 ps. step   60: timed with pme grid 104 104 104, coulomb cutoff 0.600: 6760.6 M-cycles step  120: timed with pme grid 96 96 96, coulomb cutoff 0.643: 7826.0 M-cycles step  180: timed with pme grid 104 104 104, coulomb cutoff 0.600: 6716.3 M-cycles step  240: timed with pme grid 100 100 100, coulomb cutoff 0.617: 7248.5 M-cycles               optimal pme grid 104 104 104, coulomb cutoff 0.600 step 1900, remaining runtime:     7 s           Writing final coordinates. step 2000, remaining runtime:     0 s           NOTE: The GPU has >20% more load than the CPU. This imbalance causes       performance loss, consider using a shorter cut-off and a finer PME grid.                Core t (s)   Wall t (s)        (%)        Time:      427.620      144.021      296.9                  (ns/day)    (hour/ns) Performance:        2.401        9.996 By changing just the pme_order=6, I got this: Reading file test7.tpr, VERSION 4.6.1 (single precision) Using 1 MPI thread Using 4 OpenMP threads 1 GPU detected:   #0: NVIDIA GeForce GT 630, compute cap.: 3.0, ECC:  no, stat: compatible 1 GPU auto-selected for this run: #0 starting mdrun 'Protein in water' 2000 steps,      4.0 ps. step   60: timed with pme grid 104 104 104, coulomb cutoff 0.600: 6818.0 M-cycles step  120: timed with pme grid 96 96 96, coulomb cutoff 0.643: 7821.2 M-cycles step  180: timed with pme grid 104 104 104, coulomb cutoff 0.600: 6718.4 M-cycles step  240: timed with pme grid 100 100 100, coulomb cutoff 0.617: 7257.1 M-cycles               optimal pme grid 104 104 104, coulomb cutoff 0.600 step 1900, remaining runtime:     7 s           Writing final coordinates. step 2000, remaining runtime:     0 s                          Core t (s)   Wall t (s)        (%)        Time:      550.020      144.580      380.4                  (ns/day)    (hour/ns) Performance:        2.392       10.035 I have run many test simulations by changing rlist, rcoulomb, rvdw, pme_order and fourierspacing (that I have concluded after reading papers and gromacs user list), but in most of the time, I got performance loss. I am a bit worry about using the very short rlist, rvdw and rcoulomb (most of the papers used 1.0 for each for different proteins). If I increase these cut-offs, there is performance loss, otherwise these cutoff are short (i guess). can any body guide me where I should improve the mdp file or the current one is reasonable. Thanks in advance. mayaz