[gmx-users] Lipids and virtual site in 4.6.1

Bastien Loubet bastien at memphys.sdu.dk
Mon Apr 8 11:34:49 CEST 2013

Dear GROMACS user,

Last week I was watching the GTC talk from Eric Lindahl, and I noticed his
insistence on virtual site to accelerate simulation (up to 5ps time step).

As a matter of fact our group recently tried to use virtual site on CHARMM36
POPC bilayer and the result where less than stellar.

We first did a simulation of a POPC bilayer comparing area per lipid with
and without virtual site using GROMACS 4.5.5.
We obtained an area per lipid around 0.62 nm^2 per lipid with no virtual
site but with virtual site we obtained an area per lipid of more than 0.7
After that we found that there was a bug report for the virial calculation
with virtual site in version 4.5: http://redmine.gromacs.org/issues/908

The issue being marked as resolved for 4.6 and we tried our hand at both the
beta3 and 4.6.1 version of it. The result where better but not perfect: we
get 0.65~0.66 nm^2 per lipid.

The problem here is that in order to increase the time step we need all
hydrogen to be virtual sites, so this include the lipids hydrogen and not
just the protein. However if the area per lipid is wrong the stress profile
is likely to be changed, which will affect transmembrane proteins. So we do
not feel confident about using virtual sites in our lipid simulation.

I signal it here because there is a possibility that this is still a bug, If
somebody could confirm/infirm that it would be nice.



View this message in context: http://gromacs.5086.n6.nabble.com/Lipids-and-virtual-site-in-4-6-1-tp5007063.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.

More information about the gromacs.org_gmx-users mailing list