Bas Rustenburg
Bas Rustenburg
+1 on the focus on log probability/free energy. That makes it more easy to generalize than just using equilibrium constants. We can achieve this with some refactoring. With some smart...
I don't think it'll take too long, but I agree that we should investigate first whether the rest of the code performs well enough for production simulations, albeit not optimal....
We are dealing with this as you mentioned I believe.
https://github.com/choderalab/openmm-constph/blob/master/constph/constph.py#L951
`nattempts_per_update` was (automatically/implicitly) set to 30 for dhfr, and 39 for abl, which is equal to the amount of titratable residues in each protein. As @jchodera correctly mentions these numbers...
> I wonder if just stripping the units and caching the unit-stripped values would get us most of that benefit without requiring the SWIG/Cython implementation. @jchodera I'll try it out....