Using MSMBuilder OBSOLETE

Tested on piccolo, should work anywhere.

Installation

Upon first use, issue the following command: module add msmbuilder/2.0 . If you gen an error, enable the TCL version of the 'modules' system. Add the following lines to your .bashrc

 PATH=$PATH:$HOME/bin:/shared/lab/software/bin
 export MODULEPATH=/shared/lab/software/modules/Modules/3.2.8/modulefiles/
 source /shared/lab/software/bin/module.sh

Enabling the system

Before you start any session, issue the following command

 module add msmbuilder/2.0

Trying the tutorial

To open the tutorial, call

 evince $MSMBUILDERHOME/Docs/MSMBuilder2Documentation.pdf

We shall work in a private Tutorial directory

 cd $HOME
 mkdir msmbuilder
 cp -a $MSMBUILDERHOME/Tutorial msmbuilder

 cd $HOME
 mkdir Tutorial
 cd Tutorial

For convenience, here is a list of the commands proposed by the tutorial (with minimal changes)

cp ~/msmbuilder/Tutorial/dipeptide.pdb ./
cp ~/msmbuilder/Tutorial/AtomIndices.dat ./
ConvertDataToHDF.py -s ./dipeptide.pdb -I ~/msmbuilder/Tutorial/XTC/ [--source gpugrid]
Cluster.py -r 0.03
Assign.py
CalculateImpliedTimescales.py -l 1,50 -X 1 -P 8
PlotImpliedTimescales.py -I ImpliedTimescales.dat -t 1
BuildMSM.py -l 3

ipython
> from numpy import *;
> Pops=loadtxt('./Data/Populations.dat');
> Ind=argsort(Pops)[-5:];
> print(Ind);
> print(Pops[Ind])

SavePDBs.py -c 3 -H 52 153 102 104 135 -a Data/Assignments.Fixed.h5
vmd PDBs/State*.pdb

BuildMSM.py -l 1 -w Model1
PCCA.py -M 4 -a Model1/Assignments.Fixed.h5 -T Model1/tProb.mtx -w Model1/
CalculateImpliedTimescales.py -l 1,50 -X 1 -a Model1/MacroAssignments.h5 \
  -o MacrostateRelaxationTimescales.dat -e 3
PlotImpliedTimescales.py -I MacrostateRelaxationTimescales.dat -t 1.

cp ~/msmbuilder/Tutorial/Phi.h5 ./
cp ~/msmbuilder/Tutorial/Psi.h5 ./
cp ~/msmbuilder/Tutorial/PlotDihedrals.py ./
python PlotDihedrals.py ./Model1/MacroAssignments.h5

Enabling experimental features

Issue the command (no documentation provided)

InstallExtras.sh

Local usage

Convert directories with DCD files into nested

        sdir=../filteredCA
        rm -rf dcd; mkdir dcd 
        for i in $sdir/*.dcd; do
                p=`basename $i`
                mkdir dcd/$p
                ln -s `readlink -f $i` dcd/$p/$p
        done

Using for GPUGRID

Add the --source gpugrid to the ConvertDataToHDF.py to concatenate trajectory files.

     python2.7 `which ConvertDataToHDF.py` -s ../filteredCA.pdb -I dcd -S file_dcd

Using h5 files

Command line tools to inspect binary data

Simple matrices: h5cat.py Data/Assignments.h5
Sparse matrices: mtxcat.py Model1/tCounts.mtx

Full inspection tools

> ipython
>>>  from tables import *
>>>
>>>  f=openFile('Assignments.h5')
>>>  f.root             # Shows that there is a 'Data' table
>>>  f.root.Data        # Shows details of the 'Data' table
>>>  f.root.Data[:]     # Summary of the table content
>>>  f.root.Data[1,:]   # Syntax for array access (like Matlab, but 0-based)
>>>                     # WARNING: colon syntax has a different meaning!
>>>                     # [a:b] is from a to b, b excluded (?)
>>>
>>>  ## Saving as ascii files (transpose() if necessary)
>>>  from numpy import *
>>>  savetxt('Assignments.txt',f.root.Data)
>>>
>>>  ## Working with sparse matrices
>>>  from scipy import *
>>>  tmp=sio.mmread('tCounts.mtx')
>>>  savetxt('tCounts.txt',tmp.todense())

Transforming generators into PDB files

Makes huge files!

ipython
>>> from msmbuilder import *
>>> tj=Trajectory.Trajectory.LoadFromLHDF("Gens.lh5")
>>> tj.SaveToPDB("Gen.pdb")

Updated documentation in Google Documents