2022_01_26 Shirts/Davel/Madin/Wagner - Davel next direction chat

Previous discussion (from Slack DMs)

•   Making a foyer+interchange-powered workflow to look at an interesting material/chemical question for mixed materials-organic sims

  I'm a little worried that this is won't have enough software support (on either the MoSDeF side or the OpenFF side), and it's not clear anyone will follow up on the actual science to be done on the materials.  Unless there's a killer app already ready.

 Docking rescoring (either total or just strain energies) using OpenFF. Like "run autodock vina, then erscore with openff"

I don't think this is as a scientifically high yields area.  Strain energies could potentially be valuable, but even then, if the strain is the same in solution and in protein, it's won't contribute much to the binding.  Docking scores often implicitly include "corrections" to account for the fact they are trying to match binding ensembles to single configurations (such as smaller LJ radii), so It's not clear to me that rescoring with OpenFF is automatically better. If companies are requesting this, that's a little different.

 Graph charge/post translational modification testing (not ready for 2-3 months)

Can you provide a little more information here as to the specifics?  In general - charging for posttranslational modification is very useful, so if there's a way to involve Connor in making it move forward faster that would be great.

 Alternate functional forms in the FF (would OM have to run the calcs on Lilac?)

This is important science, but it's not clear where it fits immediately in the near term to the OpenFF project - it probably wouldn't be possible to include for several years.  It might be a good think for Yu-Tang to look at, though, the timescale is a little better.

Reoptimizing BCCs against ESPs (partially started by SB and OM, progress stopped but datasets may be ready). This would be good if the new graph charges will only be calculating AM1 charges.

Yes, this could be valuable.  We should discuss how this fits into the charging plan, so maybe it would be good for Connor to attend the meeting on Friday - I doubt we would be able to say if this was a good idea or not until after that meeting.  If we DO need BCC's optimizing BCC's vs. condensed phase properties would probably be necessary as well.

• MS – Alternate functional forms is quite broad. BCC refit would be tricky, not clear that it would fit into the overall effort. Also not sure what they’d be reoptimized TO.

• OM – Yes, JW had mentioned that it’s a crowded field.

• JW – Yes, the plans for the Rosemary FF are to use a graph net to come up with AM1 charges, and then to apply SMARTS-based BCCs

  • MS – Why not a higher level of theory?

  • JW – I can’t remember the details, but Chodera/Boothroyd/Bayly talk a lot about it, they would know more.

• JW – Probably a bad idea to have CD in a production-critical role for rosemary.

  • MS – Who is doing the graph charges and refitting the Rosemary BCCs?

    • JW – Unsure, but I think SB will be starting the working group on this soon, and he may have himself/Lily/other personnel in mind.

• MS – Foyer-OpenFF simulation doesn’t seem like it’d be super high value, not sure which system would be worth simulating.

  • JW – Agree, not sure that there are interesting mixed systems that we could study, but the infrastructure should be usable with a bit of work.

  • MS – Could take monomer units and information about how they react, make a polymer out of them, and simulate something. The coordinate generation for the resulting polymer would be difficult.

    • JW – Agree.

  • MS – So, if you had a PDB that was parameterized using OPLS, could we parameterize it using openff?

    • JW – Would need to provide substructures that can be used to “tile over” the input PDB, and we could assign chemcial info to and thereby assign parameters. Would be a problem with charge assignment on that scale, until we have graph net charges.

  • MS – Could apply this workflow to polymer projects in the group, see how they behave with OpenFF force fields. We can get the monomers in SDF form, but we don’t have the polymers in SDF, so we could find a way to automatically process the monomers to make them into the needed substructures, so that we can simulate them using OpenFF.

  • MS – So this project would be “can we simulate some existing lab projects in OpenFF?”, and part of that would be generating the substructures, and possibly automating this starting from monomers.

    • OM – This could be good, and the cheminformatics would play to CD’s strengths.

    • MS – Agree, and we’re looking at some grants in this area, so this could bear fruit.

• MS will sketch out what this project would involve in more detail

• JW will send out a link to the current biopolymer/substructure tiling code