Computational Drug Discovery and Design

integration step of 2 fs. It is during this simulation that the system

starts shrinking if the initial density was not optimal:

MD 100 ps NPT at 300 K; no restraints; step 2 fs

&cntrl

imin=0, irest=1, ntx=7,

ntb=2, pres0=1.0, ntp=1,

taup=2.0,

cut=12, ntr=0,

ntc=2, ntf=2,

tempi=300.0, temp0=300.0,

ntt=3, ig=-1, gamma_ln=1.0,

nstlim=50000, dt=0.002,

ioutfm=1, ntxo=2,

ntpr=1000, ntwx=1000, ntwr=1000

/

&ewald skinnb=4.0d0 /

In the above we have added the &ewald section with one

keyword (skinnb) whose value we wish to modify. Its default value

is 2 A ̊and it corresponds to an extension of the cutoff in which the

nonbond neighbor list is created. When the system is being equili-

brated, it may shrink more than the size of this parameter. This

causes no problems if calculations are done on CPUs, but the GPU

code is not the same and may cause an execution error. To avoid it,

we altered the skinnb parameter. This value is not necessarily opti-

mal, so its fine-tuning for each specific case is encouraged. Finally,

in the final stage of the protocol we initiate a production run of

100 ns:

Production run: MD 100 ns NPT at 300 K; output once every 10 ps

&cntrl

imin=0, irest=1, ntx=7,

ntb=2, pres0=1.0, ntp=1,

taup=2.0,

cut=12, ntr=0,

ntc=2, ntf=2,

tempi=300.0, temp0=300.0,

ntt=3, ig=-1, gamma_ln=1.0,

nstlim=50000000, dt=0.002,

ioutfm=1, ntxo=2,

ntpr=5000, ntwx=5000, ntwr=5000

/

&ewald skinnb=4.0d0 /

In this case the output will be written to files every 5000 steps

of 2 fs each, i.e., once every 10 ps. This gives 100 frames of the

whole system per nanosecond. If the system is large, it will produce

154 Gre ́gory Menchon et al.