General updates | Annual meeting was a success, just about everything was prepared by new lead team, with an intro from Mobley. Annual report was sent out, governing board gave helpful feedback so Mobley is helping with revisions. The Shirts+Chodera labs released PyMBAR 4.0.0, with major improvements and some API breaks. Toolkit 0.11 is basically release-ready, working on updates to rest of ecosystem BespokeFit is being updated to check for connectivity rearrangements in torsiondrives Matt is working on finalizing some vsite/particle indexing behavior in Interchange Matt is also working on updating our ecosystem for the new OpenFF toolkit release, the main-line use of Interchange, and the PyMBAR update. David Dotson is working hard to coordinate efforts with Chodera lab and OpenFE, to get us set up to run free energy calculations on Folding@Home. For overview of efforts, see progress tracker here A lot of effort (mostly from Chodera lab and OpenFE) is going into protein-ligand dataset standardization+formatting. See PR here for details.
David Dotson is also working on preparing our QC-interfacing packages for the upcoming API-breaking release of QCFractal. BSwope has contacted us to share an analysis of FF discrepancies performed at Genentech. Will likely happen at a FF release call in September. Lee-Ping is preparing a new GeomeTRIC release - Check out this PR for the major changes. Expected to merge in ~1 day. DD – Boehringer Ingelheim had feedback for us earlier in the week - Would have preferred that we pursue Sage improvements over biopolymer development. Was wondering what this means for OpenFF. DM – I think you’re asking about “what does BI non-renewal mean in terms of feedback and our direction?”. Their feedback was more specifically that the report emphasized the biopolymer FF too much, and it gave the impression that we’re not working on small molecule FFs at all. Which isn’t the case - We’re working a lot on small molecule FFs. The non-renewal was because their management gave them the choice of supporting OpenFF or OpenFE - They chose to do OpenFE because OpenFF is already is a mature place and has usable products. TF expressed that he’s a big fan of what we’re doing and would like to support us in the future if he gets the chance. So we’ll be working on communicating “here’s what we can do with more/the same/less funding” DD – Was there also a concern from Bayer? DM – Bayer also mentioned that they’d like some revisions to the report, but they’re continuing as a partner company. MT – Sage 2.0.0 was a year ago - Even if rosemary comes out in a few months, that will be more than a year before releases (like 1.5 or 2). It looks like we have some improvements in the pipe - Do we have plans for Sage 2.1? DM – Right - PB has some good improvements lined up for the next release. Those were probably delayed from getting into a release by Simon’s bandwidth getting spread in a few directions, like graph charge research and the vsite refactor. So I’m kinda with you that I’d like to see a more continuous string of releases. MT – Yeah, it seems like there’s a lot that will go into Rosemary in the current plan. I could see us gaining some value by getting those into our partners' hands sooner in the form of sage point releases. DM – Agree, and that’s part of the revisions I’m making to the annual report. Generally, I’d say that, if Rosemary becomes delayed, we should plan to release these developments as sage point releases. JW – I’m excited that we have MT working on a new standard benchmark suite. We have a lot of rightly-placed concern about regressions in point FF releases after openff-1.3.0, and so right now there’s a lot of stress and toil around point releases. But improved stanardization+automation of benchmarks will reduce the amount of effort to cut a point release. DN – Agree. The uncertainty in these events is very expensive. If we can reduce this uncertainty and cost then this will be very beenficial for the org. DD – Are we planning to automate+generalize the previous season 1 benchmarking to establish a standard metric for releases? PB – We’re pretty good with the Lim+Hahn benchmark scripts, though we could use help automating the liquid prop benchmarks JW – … PB – I mention Lim+Hahn benchmarks because SB already made paralellized scripts to run those. So that what I’ve been using. Would be helpful to have the new automation run PL FE calcs. JW – … PB – For valence parameter improvement, are conformer benchmark improvements sufficient to justify a release? DD –For Sage generation releases, we only need to benchmark against QM conformers (energies, torsions, RMSD, etc.) and physical properties of mixtures (with e.g. evaluator). This would benchmark both valence and nonbonded parameters, respectively. We shouldn't need any binding free energy calculations to benchmark Sage generation releases (2.x). For Rosemary generation releases, we will definitely need binding free energy benchmarking via e.g. Folding@Home. JW – LW has final say on this, but improvement on conformer energies and noninferiority on phys props seems like it should justify a release. DM – PB, LW, and I had talked about this, and that’s about what we agreed on. But if the release stuff was more streamlined then it’s be easier to make a stream of small releases. MT – Given that, today, we do not have a turn-key benchmark, Plan A is “put everything in Rosemary”. DM – Yeah, and we haven’t been asked by industry to make more point releases. If they asked for it, we would. Usually our ad board is quite frank and will give us pointers about how to help them get manager buy-in PB – I think we’re pretty close to a openff-2.1 release. I’ve been looking into comparisons with the original sage benchmarking and it seems like I’ve pretty much done it. DM – DN, could you add “would you like to see an openff-2.1.0 release?” to the ad board meeting agenda? Could discuss feasibility with LW beforehand. DN – Done DD – That seems like a good plan, it does seem like pharma could use reassurance that we’re still focusing on small molecules. JW – I think this is a good idea, but we should get LW to weigh in. The cost of this could be that we don’t get eg. graph charges into Rosemary. DM – This is why I think we should discuss the possible tradeoffs with the ad board. This would fit well in the agenda. DD – Agree DN – It’s a good idea to bring this up. But it’s imporant to keep in mind that diversions in the near future can be especially disruptive, particularly given the small size of our org and the number of projects that everyone’s involved in. I also got the sense that they don’t fully appreciate the value of a consistent biopoymer FF, so this is a discussion we should have. DM – The concern from the gov board meeting wasn’t that they don’t recognize the value of a consistent biopolymer FF, it’s just that they thought that the report communicated that we weren’t spending time on small molecule FFs. So they were concerned that the report didn’t communicate to their management that we were still doing significant amounts of work on small molecule accuracy. So I wouldn’t see this as a major derailing of our current efforts.
|
