[gmx-users] Gromacs Simulation Question

[Peter Kroon]

On 07-01-17 21:12, Justin Lemkul wrote:
>
>
> On 1/6/17 5:14 PM, Academic Research wrote:
>> Hello everyone,
>>
>>
>> I have computationally designed several synthetic proteins that are
>> not found in
>> nature.
>>
>>
>> My lab has limited resources for wet lab work and so I would like to
>> use gromacs
>> to simulate these proteins in water and see weather they unfold or
>> aggregate. I
>> know the best way is to actually express and purify these proteins
>> and observe
>> them, but my idea is to use molecular simulation to screen through these
>> designed proteins and priorities the ones that appear not to unfold
>> nor aggregate.
>>
>>
>> 1. From your experiences, does my idea sound so crazy that I should
>> abandon it?
>>
>
> It's an expensive venture, in terms of time and hardware
> requirements.  To study this adequately is going to require a lot of
> long simulations.  State of the art would probably be a couple
> microseconds total for each protein, more if you can get time on an
> Anton machine (where they routinely run for milliseconds).
>
> The aggregation studies are even more expensive.  You'd have to start
> with multiple copies of the proteins in various states of unfolding,
> run replicates, and hope you can get some kind of converged sampling. 
> Having worked on small peptides (40-42 residues), this is not an easy
> task.  We did dimerization studies and a few hundred ns each was
> enough; for a larger protein I would think you'd need even more time.
>
> Assuming you have adequate computational resources, I would guess this
> is probably at least a 1-2 year project.  I've invested similar time
> on small systems.
You could also consider doing the (un)folding simulations using an
atomistic forcefield (say, CHARMM or amber), and a coarse-grained
forcefield for the aggregation studies to save on computational cost.
The latter has been done for small peptides at least [1]. Note that in
at least the Martini forcefield proteins are considered too "sticky" and
tend to overaggregate a bit, but I'm not sure how large this effect is.
There should be literature somewhere.

[1]: dx.doi.org/10.1038/nchem.2122
>
>>
>> 2. Does the Lysozyme in water tutorial from the Bevan Lab a good
>> starting point?
>>
>
> My tutorial is a simple example of how to deal with any normal protein
> in water.  Using a cubic box is inefficient, especially if your
> systems are large, so don't do that.  Force field choice is also
> critical; don't just use OPLS because my tutorial does.  Look into
> what people are finding to be most accurate for IDPs and similar
> unfolding studies.
>
>>
>> 3. Are there existing tutorials or papers that simulate protein
>> unfolding or
>> aggregation that I could use as a starting point?
>>
>
> Your simulations are functionally no different from any
> protein-in-water system, so I doubt there's a tutorial for it.  The
> logic of my lysozyme tutorial holds, and then for larger systems it's
> just a matter of placing a couple copies of proteins in a box.  As for
> literature, yes, there are hundreds of papers published each year on
> protein unfolding and/or aggregation.  Consult the amyloid literature
> for numerous examples.
>
> -Justin
>