/
Using smee to fit valence parameters

Using smee to fit valence parameters

 

Driver

Approver

 

Contributors

 

Other stakeholders

@Daniel Cole @Finlay Clark

Objective

Fit a force field using smee that yields equivalent or better performance compared to Sage 2.3.0

Time frame

 

Key outcomes

  • A force field that improves on performance over Sage 2.3.0

Key metrics

  • Primarily YAMMBS benchmarks

  • physical property benchmarks

  • free energy calculations

Status

Not started In progress Stalled Complete

GitHub repo

Slack channel

Designated meeting

Released force field

Publication

 

 Problem Statement and Objective

We wish to transition parameter fitting to using smee / descent instead of ForceBalance. The objective here is to identify a good workflow for data curation and good hyperparameters for fitting a new force field. While SPICE is an option, we want to see if we can use data computed at the OpenFF level of theory, and moreover put together a combination of optimizations and torsiondrives instead of simply single points.

 Scope

Must have:

  • A force field that performs equivalently to Sage 2.3 using OpenFF data

Must have:

  • A force field that performs equivalently to Sage 2.3 using OpenFF data

Nice to have:

  • A force field that outperforms Sage 2.3

Not in scope:

  • Nonbonded re-fitting

Project Approaches

References