Suppose, I am running a simulation of protein folding.
What are some techniques, methods, or intuitions that help me to understand whether a simulation is running properly, i.e., it is producing the correct output or trajectory?
$\begingroup$
$\endgroup$
5
-
$\begingroup$ +1 but please proofread this post and also remove the last sentence since you can either ask for techniques/methods/intuitions or symptoms, but not both. $\endgroup$– Nike DattaniJun 3 at 11:07
-
$\begingroup$ Is this question a duplicate? $\endgroup$– Hemanth HaridasJun 3 at 11:36
-
$\begingroup$ @HemanthHaridas, also remove the last sentence since you can either ask for techniques/methods/intuitions or symptoms, but not both. --- No. Nike Dattani told me a few minutes ago that I cannot ask about symptoms in the same question. $\endgroup$– user366312Jun 3 at 11:39
-
1$\begingroup$ I told you to remove the last sentence, but I didn't say to ask it again :) $\endgroup$– Nike DattaniJun 3 at 12:51
-
1$\begingroup$ If the question is a duplicate, please remove the other question. Otherwise someone else will vote to close it. $\endgroup$– Hemanth HaridasJun 3 at 13:26
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
It's a very broad question and there isn't really a clean answer that will apply for all cases but I'll just provide some general advice. I would start by re-creating a known result. Follow a tutorial for a well-studied system and make sure you can reproduce each stage of the simulation process. The most common problems with MD simulations will either cause the simulation software to crash or spit out warnings. Do not ignore warnings in the output unless you understand what they mean and are confident they are irrelevant to your case.
Visualize your geometry and trajectory! Many problems in simulations can occur when setting up your system geometry. It's very easy to make mistakes here. Visualize each stage in detail with a geometry visualization software of your choice to make sure the geometry makes sense. Don't proceed until you're satisfied that it does.
Generate radial distribution functions. These can help you catch atoms accidentally placed too close to each other. Look for any density in the RDF closer than your shortest expected bonds. Unexpected peaks in the RDFs should also be investigated.
Plot (at least) the system potential energy, density, pressure, and temperature for your simulation trajectory. Potential energy should be negative (unless the pressure is extreme), density should remain reasonable, and the temperature and pressure should remain near your set point (assuming an NPT simulation).
If simulating a protein, generate a Ramachandran plot of a series of structures along your trajectory. Dramatic changes in this distribution indicate large structural changes in your protein which may or may not be intended.
