Individual updates | MT A few weeks ago, I started updating interchange to support non 12-6 LJ, which will also be a test of interchange internals for handling other forms in general. This largely influences how some classes work. OFFTK originally had a “chain of responsibility” for ParameterHandlers in a FF modifying the Forces in an OMM System, where each one was supposed to only access specific forces. But this showed some deficiencies in order-dependence (like all charge assignment methods must finish before vsites are added).I’d copied this logic into Interchange, but it hasn’t really worked well. Right now it’s a mix of this and the factory pattern, but I don’t think it’s done evolving. As I continue working on this, the API may chance a little, but it’s unlikely to be any heavily-used methods. CC – Are custom combining rules also in scope? MT – Possibly. I’d like to know what you have in mind, and which engines you care about. CC – No specific answers yet, I’m mostly interested in alternative combining rules in OpenMM. MT – Gotcha. OpenMM will be easiest to implement this, but it will necessarily be prototype code/non-spec initially. CC – This probably isn’t an imminent need, but I’d like to get to it once I have less on my plate. MT – The timing isn’t super significant to me. My current refactor is intended to make 12-6 be more of a specific/special case rather than the only one. But in the longer run this should make it easier to feed in custom combining rules.
Also working on vsite code. Solid implementation in OMM layer, some functionality in GROMACS layer. I’m trying to move commonly-used logic outside of the writers and into some sort of common module. One decision point I have is whether OMM vsites should continue to be written using LocalCoordinateSite. This is the most general way of doing things, but it obfuscates some of the definitions if you’re looking at the output. So I may change this, and it will have an effect on the output OFFXML, but I’d validate that the energies and geometries remain the same. I may also keep the LocalCoordinateSite code around and accessible through an optional keyword. JH responded to my question about visualizing vsites, said he uses PDB. That’s good enough for me. I’m still working on internal benchmarking. LW and I agreed to put it on the backburner until graph charges are in production, since AM1BCC is the slowest part of benchmarking runs. I’m also making a nagltoolkitwrapper prototype in OFFTK. Working on cleanups in OFFTK and Interchange, with more on the way! Some chatter around ForceBalance and Evaluator. We got a new FB release out a few weeks ago, which works with new OFFTK but not with current Evaluator and Recharge. Evaluator interface was tested on a big simulation of liquid bromine. JW will handle that. Recharge API was broken by SB before he left, fixing that is longer-term. We talked with JC and LPW about software maintenance. YANK is falling behind (test suite is broken, not compatible with new pymbar) and we asked him to update, but in the next year we anticipate moving our fitting stack up to using Perses.
DD CC Out 4 weeks for UC strike, back from Dec 12. Protein FF parameter fits LiveCoMS review on datasets for protein FF benchmarks NMR benchmarks for protein FF Finished code for system setup Progress on code for production simulations After testing, will work with LWang to run production simulations on Lilac
JW –
|