We're seeing some possible methodological issues with implicit solvent
and disordered protein. Maybe some of you have seen things like this
before?
We're doing simulations of
Sic1, which is an intrinsically disordered protein that has been
studied experimentally a colleague. There is no defininet
structural model, but there is some NMR suggesting regions with
some secondary structure properties, and there is SAXS data.
There is an ensemble of structural models that has been put together
by random conformer generation followed by population adjustments to
best fit the combined SAXS and NMR data (Mittag et al., Structure 2010).
We'd like to see if a more first-principles approach free of
experimental restraints produces results consistent with these
experiments. Our approach has been to build an extended structure,
run MD and high temp (500K), then take snapshots from this run and
cool them back to lab temp (278K); and of course the hope is that this
set of low-T runs will give something like an ensemble of the natively
disordered state. However, it turns out that all the low-T runs
collapse to relatively compact structures giving an Rg that is only
about half what SAXS+NMR gives. The runs starting from different
snapshots of the high-T sim collapse to quite different compact
structures. These collapsed states are more mobile and not as compact
as regular folded proteins.
Energetically the collapse is driven by van der Waals terms. The
surface area terms also favor collapse but are insiginficant compared
to vdW. Vacuum electrostatics opposes collapse, as the total protein
charge is around +10, but the GB term counteracts this, as usual. How
the GB vs VACELE balance comes out differs on different runs, but the
end result is still collaps. The bonded terms, including dihedrals,
do nothing w.r.t. collapse.
Our implicit solvent model was igb=5, gbsa=1, saltcon=0.15 and the
default surface area parameters. MD was by Langevan dynamics with
gamma_ln=1.0. I suppose one possible problem is that our simulation
temperature, 278K, is lower than many other folks sims., and maybe
there's not so much experience with implicit solvent at that temp.
At the moment we're doing quasi-harmonic analysis to explore the
entropic side.
So does anyone have know of some experiences on simulation of
disordered proteins with GBSA models? I'm aware of Pappu's ABSINTH
model, but that's a whole other Hamiltonian.
Don Bashford
Structural Biology
St. Jude Children's Research Hospital
Memphis
Email Disclaimer: www.stjude.org/emaildisclaimer
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jun 07 2012 - 19:30:03 PDT
