Individual updates | DN KCJ tasked me with preparing the annual report. She gave me the annual report from last year. The information from the last year’s report was more specific than I can generate for this year, so I wanted to ask for your support for crafting this year’s report. Report contains info about finances and scientific/technical advances. I don’t expect you to know about financials but I do want help with the sci/tech side. Goal is to justify maintaining or increasing funding. Industry people will begin their requests for increased funding in July/August so I’d like to had off the annual report around out annual meeting at the end of june. JW – I can meet on this. Let me know if you need info from Matt or David.
DN – When do we do planning for the coming year? JW – I’m not sure. We used to do a lot of work on the roadmap but people wouldn’t be involved in planning and I wouldn’t make later decisions based on the roadmap. DN – Should we have deadlines on certain items? A target date for vsites in FFs? Or Interchange full-featured release? JW – We’ve largely avoided deadlines so far, this lets us shift priorities to do stuff like benchmarking. DN – Can we show examples of where we can demonstrate that more funding would increase our development speed? JW – The vanderbilt/MoSDeF interop collaboration took about 3 mos of MT’s time and 1 mo of my time last year, to make materials modeling functionality tht we don’t really need, and that pharma users don’t want. MT – Agree.
MT – Personally, in terms of the work I’m doing and what I want to be doing, I’m not optimistic about deadlines. I see it from your perspective that this is important at a high level and the lead team should make a decision on it. So I understand if we do start having deadlines. DN – Goal is largely to demonstrate that, if people DO want us to follow a more predictable plan, we need to be funded.
CC Prepared some new QC datasets for protein. Backbone torsiondrives for capped 3-mers. (ALA-X-ALA or VAL-X-VAL). Will be used as a validation dataset after the backbone torsions are fit using 1-mers. This is taking a long time, as predicted PB – DD deprioritized SPICE and prioritized this, so most of Liliac is likely working on this set.
An optimization dataset containing capped 1-metrs and 3-mers. Will be used for training bonds and angles. This data will be lumped together with Sage training data.
Generating librarycharges for terminal amino acids. I’d talked about doing this for mainchain AAs using LW’s work (using 5-mers). Since we also want to support terminal AAs I was able to compute those last week. JW – All permutations of ACE/NME caps, uncapped/protonated/deprotonated? CC – Yes. This has the consequence of making the caps have a neutral net charge (no charge transfer to adjacent residues). This is what CHARMM does so there’s precedent for it.
CC – With charges in place and optimization datasets ready, I’m almost ready to start fitting. So I’ll be starting doing fitting runs soon. Will do a fit where I optimize bonds and angles using the 3-mers set, and then also optimize only backbone torsions using torsiondrives Optimize backbone and sidechain torsions using torsiondrives Residue specific torsions using torsiondrives
JW – Walltime estimates for observables simulations?
MT Not much to report, focused on interchange-toolkit integration. OFFTK 1276 (Interchange integration and removal of create_force and postprocess_system) got approved last week. So now Toolkit 0.11+ will be unable to use existing custom parameterhandlers and gbsa (gbsa support will come back soon). Regression testing - I reran small molecule regression tests last week and found a single-digit percentage of times that had differences. So I’ll report on this at the regression testing meeting this afternoon. Last time this happened it took me a few days to fix, if this ends up being similar it can be fixed later this week.
PB JW – Had COVID last week, realized about halfway through Attended NIH meeting from Office of Data Science Strategy. Intersting Interesting talks of plans for legacy projects, RSE career path support. Worked on PDB loading using RDKit. Will merge this PR and then work on implementing the same functionality with OE backend. Then I’ll begin implementing support for slightly different substructure SMARTS. So if you’re making your own substructures please talk to me because the specification will be changing.
|