2024-01-10 All-hands meeting notes

Updates – Infrastructure

JW

Q4 2023 quarterly report

JE – Happy to help with JM’s workshop’s in terms of scheduling or anything else.

Updates – Science

LW

(same as above)

• JC – Loss of access to Lilac is because of xbio queue? We have a workaround now. Please contact me.

• MS – Any tests on whether the charge caching bug affected other projects?

  • LW – I talked to MT a bit about this. We expect that it should have only affected BW and LM’s work. There’s a chance that it may have affected the protein work but we don’t think that’s likely. Could test by running FB for one iterations

  • CC – We ran protein fits with Toolkit
    
    

Shout-outs

JE

• CC – Would be great to plug Jeff Setiadi's GBSA model here: Tuning Potential Functions to Host–Guest Binding Data

  • CC – JSetiadi’s been working on a SMIRNOFF compatible GBSA OBC2ish model., with radii tuned to reproduce host-guest binding energies. It shows that it’s hard to find a parameter set that works well on solvation free energies and host-guest binding free energies. Importantly, it shows that H-G binding can be a good target for fitting FF.

  • MG – Used something like 120 H-G systems to do this. Re: tradeoff between binding and solvation, the HG systems seem to over-bind with standard parameters. So we were just adjusting GB cavity radii, generally making them smaller, which improved HG modeling but made solvation energies worse. So it seems like there may be a general shortcoming with the formulation.

  • JE – Does it have anything to do with the difference between concave and convex surfaces needing different approaches?

  • MG – Good Q. I’m thinking it may be due to polarization/charge distribution of the cucurbiturils. But no proof of that.

  • JE – So it’s not a systematic mis-estimation of the surface area, but instead of the specific targets/molecules in the dataset?

  • MG – Implicit solvation models tend to overestimate the strength of salt bridges, this seems like an extreme example of that.

  •  

• MS – Polymer infrastructure paper – Work by Connor davel and a first year grad student in my group. This takes in a PDB and makes it suitable for polymer handling infrastructure, nearly accepted to JCIM. Makes it much easier to write templates for large polymers, and show that it works for NCAA, other sorts of conjugation, nucleic acids. Tools for buidling out template charges. Have folks with Micaela Matta (has a grad student collaborating with Janssen) working on polymer formulation who will continue using. Already incorporated to work with OPenFF tools. I’ll share the final paper once it’s ready.

  • JW – RIght, these will help prepare input for the custom_substructures keyword argument in Topology.from_pdb.

  • JC – Looking forward to mmcif and pdbx support.

  • JE – I suggested that some former colleagues use this to load peptoids, but I don’t think they’ve really tried yet.

  • MS – I’m happy to put folks in contact with Tim Bernat, who can probably get this working quickly.

  • (VVoelz approach?)

  • DM – I wrote the script that VVoelz uses back when I was a postdoc, so excited to see new progress on this front.

  •  

Q&A

• CC – Updates on possible change to ELF1 charge assignment in RDKit/AmberTools?

  • DM – Mixed informal results so far in testing, so not clear that this would move forward.

  • JE – Antechamber keyword implementation? Plans/requirements for implementation?

  • …

  • JW – I could just do it, it wouldn’t require a spec change, it’s just part of the implementaiton.

  • JC – I’d advocate just changing it in the implementation as a bugfix.

  • …

  • (We’ll get the report from OpenFE and if results look unambiguously good for antechamber keywords we’ll just implement it)

  •  

  •  

  •  

  •