Objective | Develop a procedure to extend a general small molecule force field to also model proteins self-consistently. | ||||||
Primary Driver | |||||||
Approvers | |||||||
Supporting Drivers | |||||||
Stakeholders (these people will be tagged in project update notifications on Slack) | Jeffrey Wagner Pavan Behara Iván Pulido Lily Wang Joshua Horton David Mobley | ||||||
Project Manager | |||||||
Page Owner (only this person can edit this page) | |||||||
Decision authority | Majority of Primary Driver and all Approvers, only in “Biopolymer FF call” meetings Veto authority: Primary Driver, any Approver | ||||||
Discussion/notification venue | Fortnightly “Biopolymer FF call” meetings (decision forum). NOT “FF release call” meetings. #ff-biopolymers channel on OpenFF slack and “FF release call” meetings (notification and discussion, no major decisions allowed here. It is not assumed that meeting attendees have read the slack discussions, they must be summarized during meetings to be considered in decisions) | ||||||
Meeting notes | |||||||
Due date | 2022-01-01 | ||||||
Key outcomes |
| ||||||
Status |
|
Overview of strategy
Generate protein QC datasets for training and validation
Fit multiple models to the same training dataset
Models vary in types of torsion parameters
All models optimize valence parameters and torsion amplitudes
Benchmark protein models
Tier 1 benchmarks for all models
Tier 2 benchmarks for models that perform well in Tier 1. Specific failures in Tier 1 may lead to new models to address problems.
Tier 3 benchmarks for release candidate
SMIRNOFF format
We need SMARTS strings that can specify protein-specific terms for general amino acids. To summarize the full discussion here: 2022-06-30 Protein FF meeting note
...
We envision several tiers of models, presented below in order of increasing anticipated effortnumber of parameter types. We will generate and benchmark lower-effort simpler models first and use the results of the benchmarks to inform decisions about higher-effort modelshow to prioritize fitting and benchmarking of more complex models. Benchmarking results for simpler models may also inspire new models not listed below to address specific benchmarking failures.
The same set of training datasets will be used for each model: Sage QC training dataset and protein QC datasets described above. Protein-specific datasets will be weighted equally for each of the twenty canonical sidechains (i.e. weights of all protomers for the same sidechain will sum to one), and the total weight for protein-specific datasets and small molecule datasets will be equal.
...
The null model is that the small molecule force field already describes proteins well and needs no protein-specific parameters. Parsley was trained on compounds that resemble protein backbone and sidechain analogs, so these parameters are likely a good first pass at describing polypeptide chains.
Amber ff99SB typed model
Backbone torsions
General backbone torsions for phi and psi
Residue-specific backbone torsions for Gly
Sidechain torsions
No protein-specific sidechain torsions
...
Backbone torsions
General backbone torsions for phi and psi
Residue-specific backbone torsions for Gly
Sidechain torsions
General sidechain torsions for chi1 and chi2
Residue-specific sidechain torsions for beta-branched sidechains (Ile, Thr, and Val)
Amber ff14SB typed model
Backbone torsions
General backbone torsions for phi and psi
Residue-specific backbone torsions for Gly
Sidechain torsions
General sidechain torsions for chi1 and chi2
Residue-specific sidechain torsions for 11 groups of sidechains from Amber ff14SB
...
Backbone torsions
General backbone torsions for phi and psi
Residue-specific backbone torsions for Gly, Pro, and beta-branched sidechains (Ile, Thr, and Val)
Sidechain torsions
General sidechain torsions for chi1 and chi2
Residue-specific sidechain torsions for beta-branched sidechains (Ile, Thr, and Val)
Beta-branched backbones and Amber ff14SB typed sidechains model
Backbone torsions
General backbone torsions for phi and psi
Residue-specific backbone torsions for Gly, Pro, and beta-branched sidechains (Ile, Thr, and Val)
Sidechain torsions
General sidechain torsions for chi1 and chi2
Residue-specific sidechain torsions for 11 groups of sidechains from Amber ff14SB
...
Milestone | Owner | Deadline | Status | Notes | ||||||
---|---|---|---|---|---|---|---|---|---|---|
Generate ELF10 library charges | 2022-06-01 |
| ||||||||
Choose benchmark systems | Biopolymer FF group | 2022-04-01 |
| In parallel with LiveCoMS review | ||||||
Generate protein QC datasets | 2022-04-01 |
| ||||||||
Fit parameters for protein-specific modelsat least two models (null and Amber ff99SB typed) | 2022-10-01 |
| ||||||||
Software for NMR observable benchmarks | 2022-10-01 |
| ||||||||
Tier 1 NMR benchmarks | 2022-11-01 |
| ||||||||
Tier 2 NMR benchmarks | 2023-01-01 |
| ||||||||
Tier 3 NMR benchmarks | 2023-07-01 |
|