Suggestions and discussion of enhancements for NAMD or NAMD support.

(Wiki dynamics favor the development of a consensus representing all points of view.)

When going through the NAMD benchmark examples listed on the "NAMD Utilities" webpage, I'm hitting the same CUDA-related unsupported NBFIX errors as previously reported on this post. Specifically, ApoA1 and ATPase are the benchmark examples that give NBFIX errors using NAMD 2.7b4 for CUDA.

If there is not a work-around, are there any plans for NBFIX to be supported in a future CUDA implementation? (The only example that I'm able to run on a GPU is ER-GRE) - JDP

The workaround is to remove the NBFIX terms from the xplor parameter file. Current CHARMM force fields do not use these terms. They will be implemented in a future version of NAMD to support Drude force fields. -JimPhillips

What about adding the capability to implement the GB/SA implicit solvent model for simulation (as in other major simulation packages), as well as additional constraint/restraint options (such as inter-atomic distance and dihedral restraints) besides fixed-atom and harmonic restraints?

Would you use an implicit solvent model? Which one should be added?

Yes. Any implicit solvent model to start with... - Shantenu

Borne? -CL

How about the ABPS module from the McCammon Group... it seems like they have a prelimiary version running with NAMD right now. Fully incorporating this in NAMD 2.7 would be nice. -JS

Would you use a polarizable force field? Fluctuating charges or point dipoles? YES! FLUC Q? -CL

YES. Fluctuation charges. - SP

Yes, especially for water molecules. NMusolino

How would you improve the NAMD interface? How would you redesign it?

What about using structure/topology files as the ones of the Tinker package? The structure file that contains the connectivity is much more clearer. Furthermore, PDB structure files are not well suited for large structures. Leandro Martinez.

How about making it so that working with AMBER or OPLS parameters is as easy as it is to setup and run CHARMM force field simulations? -CL

What kind of interface would you prefer for adding new features to NAMD? More python support. the documentation for python is weak.

What new features might be included in NAMD?

An option to periodically save the forces (total force on each atom) to a DCD file (just like the coordinate and velocity DCD files).

The Nose-Hoover thermostat.

A routine that removes center of mass motion periodically during a simulation, as opposed to just at the outset. -Neema Salimi

I would like some force field introspection features added to the minimizing routines. For instance before the first step a list of the top ten bonded and nonbonded energies and after the last step a summary of the highest energy interactions.

Additionally, a check should be added to the minimizer to stop working when the rms of forces or cosine of the gradients from the previous step and current step is small or 1, regardless of how many steps have been ordered by the user. Just suggestions.

Inclusion of the Modified Beeman integration scheme as an option. ----Brian

Some simple modifications (geometric mean for sigma) for the use of the OPLS force field. Leandro Martinez.

Inclusion of distance (optionally, angle/torsion) restraints. Pietro Amodeo

Ability to fix bond lengths (like shake) consisting of any atoms. SP

An easy way to use HIC-Up (http://xray.bmc.uu.se/hicup/) to get topology and parameters.

When debugging the problem in the NAMD Troubleshooting wiki page (A (somewhat) Repeatable Core Dump), I noticed that most .c and .C files were missing an "#include <stdlib.h>" directive. Thinking that having a prototype for malloc() and free() and friends might make a difference, even on the x86 platform, I created a patch which adds the directive to those files which lack it. Although it doesn't solve the coredump problem (that turned out to be a bug in MPICH), it does improve NAMD performance by a few percent for the apoa1 benchmark. I suspect the inline definitions which are also in that include file are responsible for the benefit.

If anyone wants the patch, let me know (tim <at> scalex86 <dot> org). I'll be happy to send it to you.