Worth Worrying About:

Ultimately, simulations are judged according to two basic criteria:

I. How well do the empirical energy surface and the chosen system composition approximate Nature?

$V(\vec{R})$:

How realistic are the chosen functional form and the associated numerical constants?

PSF Generation:

Which titratable groups should be protonated? Without employing quantum mechanics, protonations are assumed at the beginning and maintained throughout the simulation. Also, how much water is needed [17]? How many ions should be included?

II. How well is the energy surface (phase space) explored?

MD Simulation:

What length of simulation is sufficient? First, the system must be equilibrated such that system properties such as potential energy, temperature, and volume appear to have stopped drifting. Then the simulation must continue long enough to obtain reliable equilibrium averages.

MC Simulation:

Does the chosen `move set' embody all motions relevant to the question being asked of the simulation? Have enough steps been taken?

Mistakes to Avoid:

Inconsistent $V(\vec{R})$:

The potential function (long-range cutoff keywords, distances, ...) should not be changed at different stages of a simulation study. All input scripts used in a research project that evaluate energies and forces (energy minimizations, annealings, dynamics simulations, ...) should explicitly (Don't trust the defaults!) do so in the same way.

Submit and Forget:

Don't let a simulation run unmonitored. Check intermediate results daily. Plot the time dependences of the potential and total energies, the temperature, the pressure and volume (if applicable), and the root-mean-square deviation from a reference (crystallographic or $t=0$) structure.

Remember:

Simulations are fiction aspiring to emulate reality. Pretty pictures and even a few good numbers do not guarantee good science.