Hi everyone!
I'm trying to set up some aMD simulations and I have some questions. The
system I'm intending to simulate is a ~600 residues protein in explicit
solvent. I've run 100 ns of a preliminary normal MD (cMD).
According to some papers I've read (and interpreted correctly, I hope)
there may be some precision problems in the free energy map, related to the
exponential weighting [J. Chem. Theory Comput. 2010, 6, 3285-3292 3285],
which may be specially relevant for my system, given its size, and the
number of boosted degrees of freedom.
1) Being my system so big (~90,000 atoms), should I boost only dihedrals to
lower the number dispersion in the exponential reweighting factors? Will
this improve the PMF accuracy?
2) In McCamon's 2012 paper [Routine Access to Millisecond Time Scale Events
with Accelerated Molecular Dynamics; J. Chem. Theory Comput. 2012, 8,
2997-3002] they print EVERY STEP of the aMD simulation. Is this really
necesary? What printing frequency would be most efficient? (I use GPU
(GTX580) workstations with sata II 360 GB hard drives, which means sloooow
disk writing and not a lot of disk space).
3) Is it possible to combine aMD with umbrella sampling or Jarzynsky's
innequality?
Thanks a lot in advance!
Diego Alonso de Armiño
PhD student.
Universidad Nacional de Tucumán
San Miguel de Tucumán, Argentina.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Apr 06 2013 - 14:00:03 PDT
