General updates | JW – Website roadmap https://openforcefield.org/about/roadmap/ has been updated to include plans for 2023 Finishing off protein-ligand-benchmark 0.3.0 which will lead to benchmarks on F@H (which we’re now renaming) toolkit 0.14 will hopefully know how to recognise covalent modifications in a PDB Looking forward to generalised benchmarking tools that will make future benchmarking easier
JW – 0.12 OFFTK toolkit release JW – Forcebalance fork - We’ll be maintaining a minimal forcebalance fork. This will be a feature-equivalent backend for bespokefit and interface with evaluator that we can maintain more actively. This may be suitable for more rapid iteration with the fitting team as well, but do be aware that we’re not as knowledgeable as Lee Ping so there is a real possibility that bugs get introduced under our reviews. JS – Possibility of a replacement/substitute for ForceBalance? JW – It’s something we’ve talked about for a long time but I think we do finally have the capacity to handle something large like this. But it would displace a lot of other plans for 2023/whenever we do it. So at this point we guarantee that there will be a supported fork of forcebalance until we have a replaement MT JS – Would be nice to have something other than gradient descent PB – TG and I are experimenting with simulated annealing. Looks promising. Maybe more details later. JS – Will fork be conda installable? MT – It’ll be a conda package once we have it working/have cleaned up tests. ETA ~3 months +- 2 months.
LW – Sage 2.1 release. Will include big fixes that PB and CC have been working in on the past year. So updates for sulfonamides, modified seminario, symmetric parameter SMARTS. Right now PB is doing a final round of refitting, eta late jan/early feb. So, expect improvements. Also it’s been a while since a release, so this should get us back on a good tempo.
LW – nagl development, nagl and espaloma toolkitwrappers Will probably come in two forms - espaloma and nagl. In terms of the user interface, these won’t be defined as part of a force field, instead they’ll be a toolkitwrapper. So if the toolkitwrapper is installed for one of these, and you call assign_partial_charges("am1bcc"), they’ll silently be used on the backend.
DM – In terms of meetings, we’re looking at going ahead with an industry-facing meeting on May 13+14. Tentatively, internal staff retreat during May 15-18
|
