Infrastructure updates | Namespace migration – openforcefield on omnia → openff-toolkit on conda-forge OpenMM is now on conda-forge Workload planning: https://docs.google.com/spreadsheets/d/178mVhkia7UvSxMClzSwFQp4i_uvEv4kNPz1O2aUZeoQ/edit#gid=0 Industry benchmarking is going well, large “one time” expense of effort to package functionality for this specific project, users aren’t become totally self-sufficient, but are gaining faith in teh accuracy/extensibility of our tools KCJ – It’s worthwhile to do efforts like this to gain users SB – This has been causing shortfalls in infrastructure progress that I’ve needed to make up, which doesn’t seem right. (General) – May be good to dedicate a single person (or strict max number of people)to applications, and not let team effort exceed that. SB – Conformer energetics seems like a well-scoped project, but free energies aren’t too well scoped. So parts of this are a known expense, but parts will be open-ended. SB – We should make sure to refactor the existing benchmarking infrastructure to be internally useful. JW – Three phases of benchmarking Finishing season 1 (next month or two) Refactoring to make things internally useful Possibly running a season 2 around october
In person meetup after COVID? Long term/organizational
Change Zenodo metadata to have Hyesu first
Add Mitchell and D'Amore to website
Offboard Cerutti/communicate removal from services
Website insta-“cite tk/forcefield” resource
Better remote team building?
|
Science updates | SB – Preparing to answer “what comes next?” Largely studying WBO torsions. JM and PB were looking at different approaches to this starting in September. Scientific results weren’t super compelling. So WBOs won’t be in sage 2.0.0. A few lessons learned: Don’t overpromise to industry based on questionable initial results Line up infrastructure blockers earlier – we found out about systematic differences between AT and OE late in the process and don’t have personnel time to fix them before release Decide how to effectively put projects on a backburner – Both how to communicate this, and deprioritize projects such that actual progress is still made
Now that WBOs are deprioritized, JM will work on other projects, PB will start catching up on HJ’s skillset. PB, HJ, and SB are picking up the theory benchmark again. TG will work on fitting virtual sites, this could be a very profitable route. Overlaps with a lot of DC’s experience. Also looking at changing functional forms. No major progress on biopolymers. We’ve had meetings with advisory committees, but few solid outcomes. Looks like biopolymer work will begin in earnest when CC begins in June. Working with JH on making bespokefit generalize to do full FF fits. I’m refactoring fragmenter to be based on OFF Toolkit. Several toolkit blockers, some bugs that I saw in the original codebase that need resolving. Bayseian methods – Working on getting OM, TG, and TH working together on this and atom type enumeration. BCC fits are happening slowly, as I have time/resources. JW – H-G/pAPRika stuff? SB – Will probably benchmark Sage against TYK2 P-L FEs. Not clear that H-G measurements will give a lot of value. For the cost of getting more H-G data, it may be better to get more mixture properties. We’ll want to discuss gathering more physical property data since ThermoML is basically tapped out, and it doesn’t cover interesting B, F, etc chemistry. KCJ – Should we reserve some funds for getting additional physical properties information? SB – Yes. We may want to think about specifically what kinds of data we’d want for biopolymers. KCJ – A CRO would probably be better than a phd student/postdoc. SB – Could use someone cheminformatics-y on the science team to help construct datasets.
|