Fwd: Restarting T-REMD with colvars

From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Tue Nov 05 2013 - 01:48:45 CST

To be clearer, on restarting by editing everything except the shared
simulation options (as in the example provided by namd, where, however,
there were no colvars), the new t-remd starts from the same average
conformation of the peptide/protein as previous t-remd. That is my analysis
from sorted simul at the same temperature.

Hopefully I am wrong, otherwise the subsequent t-remd does not add to the
previous one. Should my analysis be correct, it would be difficult for me
to arrive at some kind of equilibration (no matter how you would like to
define it) as the supercomputer use is limited to 24h batches. Which is far
from what is needed for equilibration.

I very much hope to be noticed of a wrong procedure (with, please, a
correction as to carry out correctly a t-remd with rmsd colvars)

francesco pietra

---------- Forwarded message ----------
From: Francesco Pietra <chiendarret_at_gmail.com>
Date: Mon, Nov 4, 2013 at 5:19 PM
Subject: namd-l: Restarting T-REMD with colvars
To: NAMD <namd-l_at_ks.uiuc.edu>

I wonder whether restarting a t-remd, particularly when colvars were
imposed, goes straigth as indicated in the example of namd2.9 (or namd2.10,
which I am currently using).

That is, job1.conf (for the first restarting) reads

source fold_alanin.conf
source [format $output_root.job0.restart10000.tcl ""]
set num_runs 20000

# prevent VMD from reading replica.namd by trying command only NAMD has
if { ! [catch numPes] } { source ../replica.namd }

while the shared simulation options read:

# shared simulation options for replica exchange

timestep 1.0

structure alanin.psf
coordinates unfolded.pdb

margin 10.0
stepspercycle 100
parameters alanin.params
exclude scaled1-4
1-4scaling 0.4
switching on
switchdist 7.0
cutoff 8.0
pairlistdist 10.0

langevin on
langevinDamping 10.0

I can imagine (hope) that job1 scipt lets the simulation taking the
coordinates aftef execution of job0. However, what about if colvars were
imposed? For example, my shared simulation options read:


# Shared simulation base and options
# for parallel tempering
# System P2i implicit NVT GB-SASA

# forcefield
paratypecharmm on
parameters ./par_all27_prot_lipid.prm

# molecules
structure ./P2i.psf
coordinates ./nvt-02.pdb

# constraints
colvars on ; # provide colvarsConfig (and colvarsINput if
colvarsConfig ./P2i_rmsd-colvars.in
colvarsInput ./nvt-02.restart.colvars.state

# about temperature
langevin on
langevinDamping 10.0

# Approximations
rigidBonds all ;# needed for 2fs/step
rigidTolerance 0.000001
exclude scaled1-4
cutoff 18.0
alphaCutoff 15.0 ;# for calc Born radii default 15.0
1-4scaling 1.0

# GBIS parameters
switching on
switchdist 15.0 ;# like in manual
pairlistdist 18.0 ;# like in manual
solventDielectric 80
ionConcentration 0.15
intrinsicRadiusOffset 0.09 ;# shrinks intrinsic atomic radii default 0.09
GBISDelta 1.0 ;# calc Born radii default 1.0
GBISBeta 0.8 ;# for cal Born radii default 0.8
GBISGamma 4.85 ;# for calc Born radii default 4.85

# Hydrophobic energy
sasa on
surfaceTension 0.006 ;# like in manual

# Multistep integrator parameters
timestep 2.0 ;# 2.0 fs/step
nonbondedFreq 2 ;# nonbonded forces every two steps
fullElectFrequency 4 ;# multiple of nonbondedFreq
stepspercycle 20 ;# redo pairlist every 20 steps


My concern is at least related to the flag

colvarsInput ./nvt-02.restart.colvars.state

Is it correct to leave that unchanged on restarting? Actaully, all that
above seems to work, also restarts, however, looking into what happens
those issues above came out outstandingly. Given the cost of a 32-replica
t-remd, I would like to understand it better.

Very grateful for advice

francesco pietra

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:52 CST