Apr 15, 2020 Advisory Board Meeting
- D. Mobley provided a brief progress update since the last meeting. S. Boothroyd is working on a feasibility study for optimal combinations of physical properties for fitting nonbonded parameters. openff-1.2.0 release will contain an improved QM training set with a more systematic approach to molecule selection, while Sage will include the LJ refit based on the feasibility study. The infrastructure is moving towards providing support for WBO torsion interpolation, and SDF-compatible charges. More work is being done on protein-ligand benchmarking by D. Hahn, and J. Horton is building the bespoke parameterization pipeline. The new software scientist, M. Thompson, is looking into designing a system object.
- Another task ahead is QM benchmarking for torsions and charged molecules. F. Pickard offered advice on choice of QM level of theory for charged species.
- Suggested change for the Advisory Board officials – remove the Vice-Chair and Secretary roles with the possibility of reinstatement of these roles at a later date if need arises. K. Condic-Jurkic will draft the changes in the Bylaws and send to partners for further inspection and potentially, legal department approval.
- The in-person meeting has been cancelled due to Covid-19. It will be replaced by a virtual meeting. The format will rely on pre-recorded talks and the meeting time will be focused on discussions only.
- Two new software scientists were hired recently:
- David Dotson – Formerly of MDAnalysis, experience with MolSSI, will be working on performance/distribution for property calculations.
- Matt Thompson – Formerly of MoSdef, will be working on interoperable System format.
- Generally, increasing infrastructure capacity effectively amortizes the cost of “overhead”. Added infrastructure capacity can:
- Push out new features and bugfixes faster
- Maintain and refactor orphaned projects
- Push on benchmarking dashboard/other new or common infrastructure
- Provide support for storing reproducible calculations (like FF fits, benchmarks, etc)
- On Apr 23 the OpenFF team will have “best practices” workshop
- The Consortium could be funded for another year if the improved quality of OpenFF force fields can be clearly demonstrated – this includes QM benchmarking, binding free energies and similar, but especially improvements on drug-like molecules.
- Test sets like Baran’s “essential molecules of heterocyclic chemistry” would be a convincing set for benchmarking, but additional molecule sets could be assembled by industry partners for this purpose.
- Benchmarking dashboard might be a good way to demonstrate the improvements, but some tangible results would also be welcome (for example, predicting the correct crystal structure conformation).
The [Open Force Field Consortium Advisory Board] met on April 15, 2020. The minutes are summarized as follows: