JM just ran the first instance of the final workshop, all the materials are available here, and there will be another showing in about a month for atlantic time zones, and then there will be a directors cut uploaded to youtube afterwards.
(General) – This is the second time we’ve really gone through a deliberate roadmap-deriving process, so this isn’t a refined process but we want to make sure everyone is included.
JC – Keep an eye open for synergies with possible BTOD grant. These are items that should build on top of Consortium developments.
Proposed major objectives slide
JC – Espaloma winning would already be an OpenFF success story. We use the datasets and infrastructure.
DM – Agree
MT (chat) – How will the “Measure” bullet point be achieved? There are several details that aren’t clear on its face
JC – Can look at handling fraction of the chemical component dictionary. Then there would be two stages - first for just technical ability to load it, second for measuring accuracy. The OpenMM folks have put a good fraction of the CCD through QCArchive so that could help on the second point.
DM – And pharma is probably going to be less concerned abou the PDB per se than “pharma-relevant biomolecular chemistry” but the PDB is kind of a good starting point to help us assess, as long as we’re willing to say “but this isn’t pharma-relevant, so we’re throwing it out” etc.
MT – Do we currently not know what chemistries we fail to cover?
DM – Hemes, platinums, etc.
JE – Have we ever done a study of what % of the PDB we can cover?
(General) – No
JC – We have done datasets of druglike small molecules from varous datasets.
DM – And it shows that we do well on druglike small moles.
Next slide
(general) – Full polarizability means “direct polarizability”
JC – Audience for electrostatics improvements is internal or external?
DM – Yes
MS – Having the infrastructure in place to test polarizability would be important.
MT – Full polarizability refit in the next year? FYI polarizability isn’t in the SMIRNOFF spec
DM – Matt: At least begin working that direction. We were trying a bit for imagining an ambitious year. Yes, indeed it’s not!
MS – Yeah, I don't think anyone things full polarizability is coming this year. But thinking about how it would fit into the spec is important to start now.
CC – Willa is already fitting a water model in ForceBalance for her direct polarizability model, so there are no infrastructure blockers there.
MS + JC – Water refit?
DM – Oh, good idea. That should be in here.
JC – Custom 1-4s would be swell too, pretty sure this would give us better accuracy to QM.
DM – We felt that if we just went for “things we absolutely certainly are going to do this year” then that wouldn’t leave us in a very good place to head towards “accomplishing ambitious things on a multi-year timeframe”.
DM – Longer term things that we can’t realistically fit in here could go in the BTOD.
JC – BTOD is more for infrastructure.
JC – May be good to ensure we continue chipping away at low-hanging fruit.
MT – Perhaps it would be useful to enumerate the improvements that can be done this year to better enable << big fancy thing >> in the future
next slide
JC – When thinking about external users, it’s a big pyramid. So we need to work people to becoming contributors. And we also need stuff to be able to measure to show the NIH. So we have direct users/contributors, but also indirect users who don’t directly call out tools/FFs (like CHARMM-GUI users and OpenMMForceFields users). So should ensure that we’re supporting and growing the bottom of the force field.
DM – This could be a great area of alignment between pharma and government funder interests. Having someone come in with a FF that uses a new functional form gives us grant visibility, and also gives industry a better FF to use.
JE – Yeah, though engagement with early-stage grad students can turn into pharma users later.
JC – So, it’s important to establish routes for outside contributions.
DM – And POSE should be expanding stuff in this area/contributing to helping people getting involved, so this doesn’t all fall on Consortium fund. The workshops Josh is running are a great contribution in this area. (And attendance at those is something we can report!)
DM – And so a goal would be for this to becomes
MT – +1 to David’s distinction, and I’d add the CHARMM-GUI probably doesn’t implement out force fields correctly and I’d wager most users don’t care much about the internals (since they don’t have to) vs. first-year grad students we can get using our tools are more familiar with the internals
MT – “Have they contributed back” is the yardstick here - CHARMM-GUI none, and novice users (some-to-many)
JC – If the goal is to sell the industry consortium partners on where we’re going with this, it’s important to have a real roadmap, even if it’s speculative. Like “pay in to get vsites”, but other guarantees like “we’ll have 4 releases this year”. And if we’re aiming at modeling the whole PDB, what specific classes will we prioritize this year and in future years?
DM – Yeah, we should make a key graphic with expected releases and such.
Q&A
DM
JC – I’d like to get everyone possible involved in the RM1 process at an early stage. No concrete expectations/obligations, but I’d love to get contributions for potential people to work with (#nih-btod-may-2024 on OMSF slack)