TM Approach I: Initial Plan
- Brent Westbrook (Unlicensed)
Initial plan and approach: port to Sage 2.1 and fit to existing data
Overview
Summary | This initial approach started from the scripts and data left by Pavan, Meghan, and Jessica. Unfortunately, the re-fit force field showed worse performance overall than the original Sage 2.1.0 force field. As it turns out, many of the new parameters were not covered by existing training data, including the new torsion multiplicity data sets, so an expansion of training data and another round of fits was started in Approach II. |
|---|---|
GitHub repo/branch |
|
Status | Won't progress |
Milestones and metrics
Stage | Milestone/Benchmark | Contributors | Deadline | Status |
|---|---|---|---|---|
Re-fit 2.1.0 to TM data | Re-fit 2.1.0 to TM data | @Brent Westbrook (Unlicensed) | Jan 2024 | Completed |
Benchmark | Improved or equivalent performance on industry benchmark data | @Brent Westbrook (Unlicensed) | Jan 2024 | Failed |
Progress and findings
Industry Benchmark
The initial refit with the new data performed a bit worse than the original Sage 2.1.0 refit. To control for the presence of the new data, an additional refit of the Sage 2.1.0 force field with the new data was also conducted, which showed that the new data itself was adversely affecting the quality of the force field.
