2021-06-08 Cavender/Wagner/Behara/Pulido/Wang Check in

Current install instructions

• CC – My current workstation is windows, I’ll be getting access to a Linux one soon.

• JW – When you do have access to a linux/mac machine, you’ll follow the “developer build instructions”

  • The big difference is that you’ll want to run from a special branch, and that is topology-biopolymer-refactor

    • So run git checkout topology-biopolymer-refactor and then git pull

FF14SB port and its history

• Substructures contain entire amino acids. Since we know the space of all the molecules that can be covered by the protein force field, we can “port” it to SMIRNOFF this by generating every possible tripeptide and pulling in the parameters

  • GitHub - openforcefield/amber-ff-porting: Scratch space for porting amber FFs into SMIRNOFF format

• There’s an example of running the protein parameterization in the body text of this pull request:

• Parameterizing a protein SDF takes ~280 seconds with OpenEye backend, ~60 seconds with RDKit backend

• Removing redundant parameters from FF port should cut runtime down by another ~20%

• Torsion energies don’t always match well because of different ways to measure improper angles

Mixing force fields

• We could assign parsley valence + LJ parameters to a protein right now, but we’d need to provide the library charges from the FF14SB port, otherwise the OpenFF toolkit would try to run AM1BCC charge assignment on the protein.

  • Search in the protein FF and copy out the LibraryCharges section

Planned and existing features of biopolymer infrastructure

Kinda messy/overspecific example notebooks for new API:

Protein data files

• Merck set: GitHub - MCompChem/fep-benchmark: Benchmark set for relative free energy calculations.

• PLBenchmarks set: (The actual proteins are kinda hard to find, eg https://github.com/openforcefield/protein-ligand-benchmark/tree/master/data/2020-02-06_bace_p2/01_protein/crd )