[gmx-users] Gromacs question

[David van der Spoel]

Den 2020-02-06 kl. 15:15, skrev Kneller, Daniel:
> Hi Dr. van der Spoel,
> 
> I had hoped to reply using the user-lists but I realize now that I had 
> the user-lists in digest mode and did not want to start another thread.
CC-ing there anyway just such that it gets archived.
> 
> Thank you for responding to my question about using gmx dos and the 
> vibrational power spectrum option for gmx velacc for protein.
> I am interested in quantifying only the vibrational frequency of protein 
> secondary structure (10-60 cm^-1 ).
> In this case, would gmx dos be viable given no limitations on computing 
> resources?
> 
> I understand that in order to calculate density of states, one needs a 
> simulation with integration steps every 1-2 fs and saved trajectories 
> about every 2-4 fs. What would be considered a long enough simulation to 
> be meaningful? What would be required to show sufficient convergence? Is 
> it simply a matter of replicates or something more?
> 
 From my experience with liquids, 100 ps is enough for a liquid with low 
viscosity, but not for a liquid with high viscosity. Then there are 
typically hundreds of molecules to average over. What that means for a 
protein is hard to tell, but for sure you need 100 times the time scale 
(of 10 ps) to come up to the same level of sampling, then the protein is 
very viscous compared to a liquid, structure and dynamics have a long 
correlation time.

If you have a truly stable protein you may be better of doing a normal 
mode analysis.
> 
> Thank you,
> 
> Daniel
> 
> 
> *Daniel Kneller*
> Postdoctoral Research Associate
> Neutron Scattering Division
> Neutron Sciences Directorate
> Oak Ridge National Laboratory
> 
> 


-- 
David van der Spoel, Ph.D., Professor of Biology
Head of Department, Cell & Molecular Biology, Uppsala University.
Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
http://www.icm.uu.se