I have recently seen publications where advanced sampling methods like REMD/accelerated MD or umbrella sampling are used for loop refinement. I am particularly interested in aMD for loop refinement such as the one implemented by Song et al. (Scientific Reports 5, Article number: 7906, 2015). Do I need to impose positional restraints on the other parts of the protein while doing loop refinement using aMD or leave the entire protein flexible?
Check section S1 in the Supplementary Information of the following article. Its a very detailed protocol
http://www.sciencedirect.com/science/article/pii/S000634951400126X
I would suggest that if you are running the simulations in explicit solvent environment, then consider the rest of the protein flexible, and include the bias term only on the loop region that you are interested and run standard dynamics (unbiased) for the rest of protein. This simulations are more realistic because they consider all correlation fluctuations on the protein, i.e., also these between the loop and the rest of the protein.
The problems could arise in the case you are running the simulations in gas phase; in this case, if you consider the all system flexible, then large distortions from native structure could be expected in the rest of protein, which you may not want to have.
Hi, Neha,
I would say the opposite. If you want to refine a loop, and its starting conformation is not adequate, allowing the rest of the protein free movement could promote unnatural conformational modifications in it. Using positional restraints in all the protein or in its backbone, and allowing the loop free movement seems to me a better way to sample the loop conformational space, and then picking up a proper starting point for a conventional MD (how to pick up this conformation is another point, perhaps a most abundant state).
I do not see any meaning in running MD simulations of proteins in vacuum, but you could consider explicit or implicit solvent. And the choice here is not trivial.
With a long loop, the conformational sampling from MD could not be sufficient, and could be difficult to properly characterize it as adequate. You could use first some loop optimization function prior to the MD dynamics. Also, to increase the sampling speed, use larger integration steps, such as 4 or 5fs, associated with v-sites (at east in GROMACS) could facilitate. A possible way to analyze the results could be a free energy representation of the sampling, to pick up a more stable conformation.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation