Individual updates | CC Most of last week was working on automated protein benchmark job submission Solved a pretty big hurdle - Having OpenMM model protonated cterm ALA. This is because the NMR data was gathered at a pH of 2. It seems like things are going well now. So only remaining hurdle is automating the process of submitting script to queueing system. Continuing to work on LiveCOMS review. Implementing form changes that MGilson and I discussed. Need to revise draft and get it sent back to authors. Changes are mostly figuring out where we needmore details in order to standardize the sections/voice, especially in terms of how simulations are compared to experiment. Meeting with Janssen team about biopolymer support (during core devs timeslot next week). Hope to provide broad timeline, and understand what they can do to support our efforts. I’ll present a draft of what I’ll say at the biopolymer ff meeting on Thursday. JW – DN and I can work together with you on a timeline for this - We’re wondering when we should aim to have the F@H infrastructure ready for protein ligand calculations, and when to plan for rosemary finalization. CC – I’ll message you two and we can find a time to meet about this before Thursday.
MT Last week I implemented “refactored vsites” in interchange and it was way easier than anticipated. Simon’s refactor was great, but targeted the 0.10.X branch. When we ported that PR to the 0.11 branch, there were certain pieces of functionality that couldn’t be directly brought forward in the toolkit, since the new home for the logic was in Interchange. So we took some major tests that couldn’t be forward-ported in the Toolkit and added them as requirements in interchange. I took over control of openff-evaluator - There hadn’t been a release for a while, and it wasn’t clear who had which privileges (PR approval, new features, releases, etc), and I’m going to cut through that. So right now I’m pruning Evaluator down to 2 “Real” branches - The “0.3 line” which will remain compatible with the Toolkit 0.10 line. Then there’s an “0.4 line” that will start getting prepared for the toolkit 0.11 line, which won’t be released until the toolkit 0.11 comes out. I can run an RC period for that if folks want. Generally when I take over a project I try to reduce the number of open issues and PRs. I figure this is what we want - When users want a lot of things, there should be a small number of open issues/PRs. If there are too many/too few open PRs/issues then things aren’t idea. I generally keep tests running for the development head, NOT the last released conda package. This means that live conda packages may encounter dependency breaks or something, but that we can always quickly cut a new release to fix that. I will focus on maintenance, not new features. So if there are major developments needed, I’ll need to hear about that and figure out what my role is. DN – I agree with not committing to new features by default. New features should be tracked on the team backlog, and new items should be discussed and agreed-upon with the team before work begins. MT – I agree with DN. I think there’s some amount of unknown new developments in the pipe for evaluator, and we should discuss this with Lily to understand if we’re expected to do some of it or what needs to be prioritized.
CC – For current work, should I use the 0.3 evaluator line, or the 0.4?
Merged a lot of toolkit PRs. One big thing is that we’re dropping the “old” create_openmm system code path completely, and all of the logic now go through interchange. The old code path is now completely inaccessible and will soon be removed. So, a few thousand lines are being deleted from the toolkit. Dealing with some very niche edge cases now to make tests happy and match old behavior. This is largely my “finish line” for the initial interchange release.
JW Bespokefit pre-release testing and polishing - Aiming for release ASAP once quick start guide is a bit more polished. DN – I saw that the blog post PR was merged last week. Will the full “release” be announced soon? It seems weird to have the blog post out but not be ready for users to start trying out the release. JW – This was my mistake. I think we can keep the blog post up, since I’m hoping to announce the release today. DN – Agree
Got some good feedback from OpenEye about proteins that aren’t loading, compared loading them using from_file and from_openeye. Will work on that this week. Largely catching up on backlog. Toolkit under 20 PRs (big thanks to Matt). Some tech support. RGowers may have found a faster way to load proteins! Uses chemically invalid molecules and RDKit SMARTS matching.
DN Some F@H project planning, looking at figuring out deadlines, but we should probably do this tomorrow while DDotson is available. Generally, we should figure out timelines for the F@H project - I’ll chat with CC and JW about this later at a separate meeting.
|