I work with peptide-bilayer interactions and I have tried a few strategies to observe the partitioning and folding of peptides, however without good results. If the problem is sampling, Replica Exchange MD has been suggested as a option for improving the sampling, but with precaution of the results. If the problem is time, I have worked with CG to study peptide-bilayer for a few microseconds, but or the FF have problems with helix formation or I should let it running for milliseconds, which is impratical.
Comments in this topic from people who work with these models could be enlightening for me.
Hello Bruno:
Indeed, observing the folding of a peptide can become a non-trivial problem to simulate. The free energy landscape is often populated by multiple local minima with different energy barriers, so it can take you time to sample exhaustively, overcome those barriers, and manage to sample the conformation of interest. In addition, the peptide in aqueous solvent may have an A conformation (e.g. random coil). In contrast, when partitioned to the membrane, it prefers a B conformation (e.g. an alpha helix), which increases the complexity of the problem.
The option of improving your sampling sounds appropriate, perhaps using other flavors of replica exchange, such as REST2 or well-tempered metadynamics, will help you overcome the energetic barriers more easily. You might consider using a collective variable that better describes the folding, e.g., number of contacts and RMSD, you might have more success with that.
A key thing here is the force field you are using, you don't mention which one you use, but you should know that some "over stabilize" certain secondary structures compared to others (e.g. some amber flavors are very prone to form helices).
Good luck!
Thank you very much, Mr. Otálvaro.
The forcefield is a thing for me too. I have tried to reproduce an experiment described in this work: Peptide Partitioning and Folding into Lipid Bilayers (10.1021/ct900256k), figure 5, and, because of the DMPC bilayer, I have found easier start using CHARMM 27 ff. The peptide begun to fold in solution (I have NEVER seen such thing before). So, I changed the ff for GROMOS 43a2, as described in the paper, and the reverse happened: no folding (1 microsecond of simulation). I do believe I jumped some step of the paper methods, but this summarizes very well my Don Quixote journey on peptide folding.
You do mention that some FF are more prone to fold structures than other. The question that comes up is: how reliable are these results?
Hello Dr. Bruno,
As far as my knowledge on these topics goes, I would recommend the following readings:
- Molecular Dynamics Simulations Are Redefining Our View of Peptides Interacting with Biological Membranes (10.1016/j.sbi.2019.12.021).
- Understanding and Modelling the Interactions of Peptides with Membranes: From Partitioning to Self-Assembly (10.1016/j.sbi.2019.12.021).
- Computed Free Energies of Peptide Insertion into Bilayers Are Independent of Computational Method (10.1007/s00232-018-0026-y).
Also, as Mr. Edward mentioned, the free energy landscape is often populated by multiple local minima with different energy barriers. In this context, molecular dynamics simulations of peptide-membrane interactions are regularly hard to study. I personally have tried High-Temperature MD, which Professor Martin B. Ulmschneider has reported in various studies. I have achieved good results in peptide folding and also in peptide insertion; nonetheless, these results should be supported with experimental data.
Best regards,
Zuriel
Zuriel González Carrera, thank you for your response. I will take a look at the studies you have mentioned.
Also, you have mentioned to have obtained good results with HT-MD, correct?
Could you tell me which forcefields you have used in these assays?
Hi Bruno:
What you describe is not surprising: The free energy landscape describing peptide folding is surely populated by multiple local minima with energy barriers of different magnitude, so you may be jumping into a local minimum with a high energy barrier that you cannot escape from on your simulation timescale.
If you want to reproduce the results of the article you must do as you say, use the same simulation parameters and force field.
Talking about reliability of results then falls back to statistical significance: Is the folding you describe just an anecdotal case or do you have multiple replicates that allow you to build a statistic about the reliability of your result? This is of course without taking into account what you have also mentioned: Have an idea of the experimental behavior of what you want to simulate. Try to look at the most recently parameterized force fields for your system of interest.
Regarding high temperatures to improve sampling: Be careful that by raising the temperature and then decaying, you don't end up with a local minimum that is then difficult to escape from and is far from the conformation you want to sample mainly. You could try the methods I told you before, they have worked for many people before.
Okay, I am starting to see a pattern here for forcefields:
Article Absorption and folding of melittin onto lipid bilayer membra...
: Figure 1 elicits the partitioning and folding of a previously unfolded peptide in a POPC bilayer under 2 microseconds simulation at 120°C. The forcefield is CHARMM36/AA.Article How reliable are molecular dynamics simulations of membrane ...
: Figure 4 compares the peptide folding in three FF: CHARMM, OPLS-AA and GROMOS96-53a6. The highest percentage of folding is CHARMM. The authors discuss that this difference is because the used CHARMM lipid as AA and the the FF as UA. Temperature of 35°C. Time of 1.5 microseconds.
You can include this video and the references therein:
https://www.youtube.com/watch?v=AtDOJnVNC18&ab_channel=softsimu
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs