Sep 17, 2020 Advisory Board Meeting
Posted on 17 Sep 2020 by Karmen Condic-Jurkic
The Open Force Field Consortium Advisory Board met on September 17, 2020. The minutes are summarized as follows:
Funding and personnel updates
- K. Condic-Jurkic provided an update about funding targets for Year 3 of the OpenFF Consortium.
- D. Mobley announced a successful application for the NIH supplement grant for improved code interoperability, led by M. Shirts. Feedback was requested about the key interoperable features desired by the pharma partners. At the moment, building an interoperable system object is in the focus of this grant.
- A new team member, Pavan Kumar Behara, joined in September 2020 to work on the QM related parts of the project, such as data generation and validation pipelines, and force field optimization.
Minor Parsley bugfix
- This bugfix release manually changes two bond force constants (
b24
andb27
) to resolve an issue seen in propyne substituents when using hydrogen mass repartitioning with a 4 fs timestep. Full details are available here and in thereproduce_propyne_hmr.ipynb
notebook in the release assets on GitHub. More systematic solution expected in the future.
Industry-led benchmarking study
- G. Tresadern (Janssen) suggested an industry-partner led benchmarking study which would take a closer look at OpenFF performance in comparison to other force fields when applied to proprietary data, based on the Lim et al study. Participating partners would run their own assessment and report the results back without revealing the molecules used in the study. The proposal has been accepted and will be led by Janssen.
- This study would be released ahead of more complex reports on protein-ligand free energy benchmarks, as it would be easier to interpret in terms of performance. Some of the areas of interest for this study include:
- Applicability domain (parameter coverage)
- Accuracy (energy conformers, energies etc)
- Outliers
- Testing the available tools (Evaluator, Toolkit, conformer notebook)
- Each partner will need to adapt protocols for local installation and deal with dependencies, which might require some testing. An OpenFF person will be assigned to assist with tools and protocols required by this study.
- Next steps:
- List all interested volunteers from industry
- Janssen will lead the study, start refining the formal benchmarking protocols in collaboration with OpenFF and run early tests
- Every partner should test the workflow before running the official study to make sure that everything works as intended.
- Additional suggestions for the study:
- Investigate torsional drives and barriers
- Comparison to ANI in addition to QM and other force fields
- Identify some of the key functional groups that might be problematic / have a poor performance.