Notes

The transition from pDynamo2 to pDynamo3 should be quite straightforward as the scripts are quite similar, apart from the changes required by going from Python 2 to Python 3. However, the underlying structure of pDynamo3 is substantially different and its package system has been radically overhauled. Users should be able to get a good idea of how to modify their pDynamo2 scripts by comparing equivalent examples in the pDynamo2 and pDynamo3 distributions.

This a list of some of the features that it be would be nice to have in pDynamo. It is based on a summary that was discussed at the Barcelona workshop in September 2012, and which has subsequently been updated to take into account the more recent pDynamo changes.

• Check-pointing. This is straightforward and can be done readily by users in their own scripts. A more general mechanism could be implemented but needs to be defined (i.e. what is to be check-pointed?).

• Constant pressure molecular dynamics. Using a Langevin method.

• Documentation. This is unlikely to happen and users should rely on the community, the examples and tutorials in the distribution and the wiki.

• Ewald/PME methods. This was done by Nicolay Simakov in the group of Troy Wymore but needs to be incorporated into the release version of pDynamo. The problem is to make it consistent for use with MM, QC and QC/MM methods.

• Implicit solvent methods. For example, GB/SA (as in fDynamo) or something equivalent. Unfortunately, there seems to be no single existing method which is satisfactory for MM, QC and QC/MM methods.

• Inter-image bonds. For the simulation of covalent crystals, infinite polymers, etc.

• MM polarization. Feasible but very unlikely unless good, parametrized force fields become available.

• Parallelization. Parallel versions of fDynamo and pDynamo2 existed but a concerted effort on pDynamo3 has not yet begun.

• PDB file handling. This is a constant effort but radical changes are unlikely.

• Regular developer and user meetings. A good idea but it depends on funding!

• Sampling algorithms. There are many that could be implemented, including a docking algorithm (c.f. with the QuickDock algorithm in fDynamo) and metadynamics (available in external programs which could be linked to pDynamo).

• SCF convergence. Another constant effort (not just for pDynamo!). A full second-order method would be nice although take an effort to implement.

• Units. The standard units in pDynamo are Angstroms, atomic mass units, kilojoules per mole and picoseconds, except for purely QC energies and properties which employ atomic units. When non-standard units are used, they are usually (and should be) specified.