CFACV -- a collective variables implementation for NAMD

Cameron F Abrams, Drexel University, Philadelphia, Pennsylvania
cfa22@drexel.edu
215 895 2231
(c) 2016-2018

CFACV is citable and archived at Zenodo:

Please also cite "Cameron F. Abrams and Eric Vanden-Eijnden, "Large-scale conformational sampling of proteins using temperature-accelerated molecular dynamics," Proc. Natl. Acad. Sci. USA 107 4961-4966 (2010)"

INTRODUCTION

CFACV is a simple collective variables implementation for NAMD that uses tclforces and implements Temperature-Accelerated Molecular Dynamics (TAMD). Collective variables are functions of atomic Cartesian coordinates, and CFACV currently allows for CV's to be centers of mass of groupings of atoms and any lengths, angles, and dihedrals constructed thereof. These CV's can be harmonically restrained or driven with Temperature-Accelerated MD.

In this release (v1.0), there are three subdirectories, summarized below:

1. src -- C files and headers, plus a makefile
2. tcl -- tcl scripts
3. examples -- three example TAMD simulation inputs

INSTRUCTIONS FOR RUNNING THE HIV-1 MA EXAMPLE

1. Compile shared object libraries cfacv.so and libgenericdataspace.so

   From within the main directory:

   • mkdir lib (you only have to do this once)
   • cd src
   • make cfacv.so
   • make genericdataspace.so

   Note that you must have swig installed.

2. Run the example (This is a system with HIV-1 MA (PDB 1hiw))

   • cd into examples/hiv1-ma
   • examine mk_tPDB.tcl; you can edit parts of this
     Currently, this is set to block the protein into 5 subdomains
   • issue "vmd -dispdev text -e mk_tPDB.tcl -args -p my_system.pdb" This will create "label.pdb"; "label.pdb.BAK" is a copy I created.
   • issue a command to launch namd2 with the config file "example.conf" For example, charmrun ++local +p4 /usr/local/bin/namd2 example.conf +idlepoll
   • watch it run -- the log file has lines that begin with "CFACV/C) that report data on fictitious variables "Z" and congruent collective variables "Th" (for "theta")

The example in the directory alad/ is CHARM22 alanine dipeptide (residue ALAD) in TIP3P water, with the phi and psi angles as CV's, driven by TAMD. The directory ala2-fixed-phi-psi/ is an example in which cfacv is used to simply restrain at a particular (phi,psi).

INSTRUCTIONS FOR INSTALLING CFACV

Prerequisites:

1. swig and tcl headers and libraries

2. Gnu Scientific Library

To compile:

1. copy the cfacv/ directory wholesale under your home directory

2. cd ~/cfacv ; mkdir lib bin (if you haven't already)

3. cd src; make all

Note that this uses swig to build a shared-object library that the TcL interpreter in NAMD will load. For this reason, it is a good idea if the TcL version that swig uses matches the TcL version that is "inside" NAMD. If there is a mismatch, a good way to fix it is to compile NAMD from source using the system TcL header and libraries rather than the TcL the NAMD compilation instructions suggest.

In any run directory where you want to use cfacv, make sure the NAMD configuration file references the right path for cfacv_tclforces.tcl as being $env(HOME)/cfacv/tcl/cfacv_tclforces.tcl or wherever you chose to put this repository.

CAN I GET THIS TO WORK ON STAMPEDE2?

Yes! Jim Phillips' most recent build of NAMD 2.12 (as of this writing, /work/00288/tg455591/NAMD_LATEST_Linux-KNL-MPI-smp-Stampede) cannot use CFACV because of a TcL incompatibility. However, there is a version of namd2 in my Stampede2 directory that does work with CFACV, and it can be found at /home1/00634/tg457991/namd/NAMD_2.12_Source/Linux-KNL-icc/namd2.

MORE DETAILS

CFACV interfaces with NAMD via a tclforces script called cfacv_tclforces.tcl that initializes a workspace and then defines the all-important calcforces procedure. Procedures called from this script are defined in cfacv.tcl and much of the implementation is housed in cfacv.c. Swig is used to interface cfacv.tcl and cfacv.c to create a shared object library 'cfacv.so' loaded by cfacv_tclforces. (Note: you need swig installed to build the shared-object library.)

The configuration file 'example.conf' contains a tclforces stanza for a TAMD simulation. In addition to defining the tclforces script, the TAMD implementations need the names of three accessory files:

• labelPDB: this is a PDB congruent to the simulation system in which the beta field of each atom designates its membership in a "group". (0 designates the "null" group.) You have to create this file yourself. The mk_tPDB.tcl script illustrates how one can use automatic clustering of residues into subdomains to generate subdomain groups. Groups can have a minimum of one atom.

• cvINP: this is a custom input file that identifies the types of collective variables and what groups they involve. The example cv.inp in this directory defines the cartesian coordinates of the center of mass of three groups as six collective variables. There are other types of CV's; consult the source.

NOTE: the group ID explicitly listed in the beta field of the labelPDB is an index that begins at 1. "1" in this context means "group 0", the first group. Groups listed in the cvINP file indexed beginning at 0. So, all atoms with a "1" in their beta field in the labelPDB correspond to "group 0" in the cvINP file; those with "2" in their beta fields are "group 1", etc.

• restrINP: this is a custom input file that defines how the CV's are restrained and/or driven by tclforces. It has a matrix definition of the restrained variables (which are generally linear combinations of CV's), and each row also lists TAMD parameters in a self-explanatory way. NOTE: this input file is optional; if it is not listed, you must have a line in the NAMD configuration file that looks something like

set restrPARAMS {{k 100} {TAMDkT 6.0} {TAMDgamma 100.0} {TAMDdt 0.002}}

In either case, time is measured in ps for all TAMD variables.

To understand the implementation, I recommend tracing the code beginning with cfacv_tclforces.tcl. Issue make cfacv.so to build the shared object library.

If you are only going to accelerate Cartesian variables, this is good enough. If you are going to accelerate dihedral angles, you have to indicate that the restraint function in periodic with the additional TAMD parameter {rf PERIODIC}.

Please contact me with any questions or problems. This is not a professional software release, so I fully anticipate there are bugs or other confusing issues.

ACKNOWLEDGMENTS

1. VMD and NAMD are products of the Theoretical and Computational Biophysics Group at the NIH Center for Macromolecular Modeling and Bioinformatics at the University of Illinois at Urbana-Champaign

2. All codes and data in this repository have been made possible with partial support from NIH through grants AI084117, AI093248, GM115249, GM056550, and GM100472, the National Science Foundation through grants DMR-1207389 and MCB-1330205, and the US Army through grants W911NF-12-2-0022, W911-NF-13-1-0046, and W911NF-12-R-0011.

2017-2018, Cameron F Abrams

cfa22@drexel.edu