A slight distinction needs to be made here. The protein folding from 1D through to 3D structure is a millisecond-scale process whose atomic-detail simulation is extremely difficult using current capabilities. However, if you aim to predict the final folded 3D structure of a given sequence, then the protein structure prediction, especially the homology modeling tools, are the stuff you need to apply. At the next step, you may refine the predicted structure via MD simulations.

A slight distinction needs to be made here. The protein folding from 1D through to 3D structure is a millisecond-scale process whose atomic-detail simulation is extremely difficult using current capabilities. However, if you aim to predict the final folded 3D structure of a given sequence, then the protein structure prediction, especially the homology modeling tools, are the stuff you need to apply. At the next step, you may refine the predicted structure via MD simulations.

No! Using classical MD simulation, you can not predict the folding of an unsolved protein; however you may use Monte Carlo (MC) methods to roughly study about it, You can start using Rosetta protocols:

https://www.rosettacommons.org/software

D.E. Shaw has used his MD program, Desmond, and dedicated computers, Anton and Anton 2, to fold linear peptides into 3D structures. However, for proteins whose 3D structures are unknown, you could try homology modeling or the procedures of protein structure prediction embodied in the I-TASSER suite. The quality of predicted structures can be assessed using tools included in I-TASSER or Molprobity.

I agree with other answers that predicting 3D structure requires homology modeling due to the fact the MD can't do it in reasonable time. Another possible approach is to run MD with coarse grain model.

If you were able to fold even a 50 residue protein in a given solvent using MD you would be first rank

candidate for Nobel prize.As the other colleagues suggest above homology modeling maybe helpful

provided there are enough known protein structures with homologies>25% to yours.

Sandeep,

I would like to second Rudy's suggestion above. Starting from a protein sequence, I have used I-Tasser (https://zhanglab.ccmb.med.umich.edu/I-TASSER/) and Modeler (https://salilab.org/modeller/) as the starting point for MD protein simulations. If you know of a similar molecule with an NMR or X-Ray structure, that helps. It is easier to guess the 3D solution based on homology and see if it is stable than to try to fold it from scratch. In the cell, there are many other parts of the folding machinery besides solvent and your protein sequence. You can't expect it to simply happen without this complex environment (at least not in a reasonable amount of time).

Yes, I agree, if there is a crystal structure of a protein from the same family with at least %20 amino acid sequence identity, comparative methods known as homology modelings can be useful to construct an initial conformer.

Thanks all for valuable suggestions, this is a ~1300 aa long secretory protein from a bacteria secreted by type V secretory system and having a 33 aa long signal sequence and an autotransporter domain. Protein matures in periplasm and then either secreted or decorated on outer cell membrane. This protein has sequence diversity from different strains of same species and only some small segments have 10-20% structural homology with other characterized structures. For this protein, structure is solved only for a small segment ranging from 388-844 aa which have no similarity with other proteins. Multiple sequence diversity divides this protein into different clusters. Would it be relevant to create consensus sequences for these clusters, modeled them and then simulate.

Hi Sandeep

Can you send me the sequence and pdb file of known segment?I would like to study it a bit with my

own tools and maybe provide some feedback.

Dikeos Mario

Hi Sandeep

i am working on it.This protein is a nightmare for all predictors I know.But

I will send you a rough initial structure in pdb format soon.

Dikeos Mario