I wanted to have a rough estimate on time it takes to get the complete thermal distributions of backbone Ramchandran angles of a given peptide size of 18. Is it possible to estimate the time before doing the Monte carlo simulations. In the attached paper they have mentioned it took 108 CPU hours with 30 cores for 200 milliion CPU steps for TRP cage 20 residue protein. With this info is it possible to estimate the wall time of the MC simulations of a given biomolecule ?
https://www.nature.com/articles/s41598-020-75239-7
In my opinion, that time will be sequence dependent, since not all the residues have the same regions available in the Ramachandran angles space. After all, Monte Carlo is a sampling methodology, and the "correct size" of a statistical sample is always system dependent.
When I use MC, I always run several independent samplings (using a different seed number for the random number generator). If the results are statistically equivalent (according to you accuracy criteria, system dependent again), I assume that the individual samplings are correct. In addition, you can average over the results of the independent samplings. I prefer this strategy, with several medium size independent samplings, to one individual very large sampling, where you may have become trapped in metastable states.
- Badges
- Science topic
- Similar topics
- Engineering
- Modeling