REMD with AMBER

From CUC3
Revision as of 10:06, 14 July 2008 by import>Bs360
Jump to navigation Jump to search

To run REMD with AMBER you need 'sander.MPI'. You can either copy it (clust:~bs360/bin/sander.MPI) or create your own. To do so you have to copy the directory clust:~bs360/MPIAMBER/ and follow the instructions on http://wwmm.ch.cam.ac.uk/wikis/ucc/index.php/Notes_on_dove#64-bit_AMBER9 to create the config.h (or take the existing one in the MPIAMBER/src/ directory and change the path names accordingly). After producing the config.h file you can compile the MPI version by 'make parallel'. You should have loaded the pgi64 and mpi/lam/64/pgi/7.1.1 modules.

To run REMD with AMBER you need following input files:

-> 'numreplicas'

  (number of repilcas)

-> 'groupfile'

   -A -i mdin.rep1 -o mdout.rep1 -c inpcrd.rep1 -p ./ala2.prmtop -r restrt.rep1 -x mdcrd.rep1 -inf mdinfo.rep1
   -A -i mdin.rep2 -o mdout.rep2 -c inpcrd.rep2 -p ./ala2.prmtop -r restrt.rep2 -x mdcrd.rep2 -inf mdinfo.rep2
   ...

Here, mdin.rep*, inpcrd.rep* and ala2.prmtop define the other input files; all other files are for output.

The inpcrd.rep* and ala2.prmtop are the usual coordinate and topology files and mdin.rep* define the input for the MD for each replica:

REMD run 1

&cntrl
       imin = 0, nstlim = 500, dt = 0.002,
       tempi = 100.0, temp0 = 100.0,
       ntt = 3, tol = 0.000001, gamma_ln = 5.0,
       ntc = 2, ntf = 1, ntb = 0,
       ntwx = 500, ntwe = 0, ntwr =500, ntpr = 500,
       cut = 99.0, igb = 0, saltcon=0.0,
       nscm = 500, irest=0,
       ntave = 0, numexchg=100000
&end

An exchange is attempted after each cycle. The cycle lenth is determined by (ntslim)x(dt), i.e. (MD steps)x(step) length. ntwx determines after how many MD steps the coordinats are written to mdcrd.rep*. ntpr determines after how many MD steps the energy is written to mdinfo.rep*. ntwr determines after now many MD steps the restart files mdinfo.rep* is updated. The temperatures for each replica you have to determine yourself, i.e, they are not automatically exponentially distributed between T_min and T_max as in the MMTSB protocol.

You can find an example for REMD with AMBER in clust:/sharedscratch/bs360/amberrex_test/