2021-03-12 LW/JW/DM Meeting notes

Overlapping project ideas

• (JW and LW had previously discussed LW/OpenFF infrastructure interests on Slack)

• (General) – OpenFF isn’t going to be providing polymer building tools, but we’re very interested in reading polmyers form common formats, percieveing important things about them, and assigning parameters.

  • Requisite – Handle branching

  • General integration of polymers into SMIRNOFF format?

  • Grafting not-polymer stuff onto polymers?

  • Further psiresp development+integration?

    • JW – This would be cool – Could plug this in to OpenFF toolkit (ie calculate the charges for this molecule using RESP)

    • LW – May be interesting to use this in the case that the final wavefunction is already available.

    • DM – Considerable interest on our side to study methods that can get good “quality” charges without doing full QM each time.

  • Bespoke polymer FF creation?

    • Input:

      • Individual units and their connection points

        • HELM/BigSMILES

          • How to specify different compositions/defects/terminating units?

      • Or a single instance of a large complex polymer, and automatically chunking it into smaller units, then making a polymer FF based on those units

    • Output:

      • A force field (OFFXML)

    • Technical developments:

      • Atom type explosion

        • LW – Could define residues so that, if they would get different junction parameters because of their neighbors, they’ll be a different residue

        • JW – That should work, we’ll just want to make sure the final “type assigner” knows about this.

      • Enumerate possible building block connections

        • Should one building block’s parameters potentially be affected by the building blocks adjacent to it?

          • Could start with a plugin for heuristics, starting with “everything out to one building block at each connection point”

        • Tools for enumerating linkages:

          • HELM

          • Reaction SMARTS/SMIRKS

          • BigSMILES (But not supported by many packages)

          •  

• Coarse graining?

  • DM – Trevor Gokey in my lab is interested in this, but it differs significantly from OpenFF’s existing work.

• How to do substructure searching?

  • DM – It’d be great to use this using chemical SMARTS. Once molecules go into atom-type land it’s hard to recover them.

• LW – I’ll eventually be doing torsion drives

  • JW – May be useful to look at protein torsiondrives

• DM – Stuff that OpenFF contributes to has to be MIT or BSD-3. This allows commercial users to use our stuff, which is different from GPL. We’d be very interested in helping replace any GPL-dependent code in your workflows.

•  

Next steps

• DM – Basically we want to figure out how your existing plans overlap with what OpenFF can offer, another is to identify a new research direction and focus on that here.

• LW – Would consider working past 9 months if things progress well. So “new project” is an option. Bespoke polymer infrastructure would be cool.

• DM – We’d be interested in either the “here’s a complex polymer, cut it up and make me a FF” functionality. We’d also be interested in smarter “something that isn’t in the polymer FF is stuck to a polymer, how do we graft it on?” logic, including smart grafting rules, or quality estimation.

• LW – Would be interested in turning over maintainership of the final project. Not tied to parsnip name, would be happy to fold this into OpenFF tooling.