| Excerpt |
|---|
Constrained fit to initial values with additional Br data |
👀 Overview
Summary | This fit adds additional Br ESP data following a hypothesis that a lack of Br data was responsible for the poor performance of Br compounds in condensed phase properties. Performance unfortunately did not improve. While working on this approach, it became apparent that we weren’t sure where the initial values for each virtual site came from – while they had seemed to come from Qubekit, we could not reproduce them in Qubekit. In Approach V we took a step back and restarted the project from Qubekit initial values, as well as exploring other directions. | ||||||
|---|---|---|---|---|---|---|---|
GitHub repo/branch | |||||||
Status |
|
| Table of Contents | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
\uD83D\uDEA9 Milestones and metrics
Stage | Milestone/Benchmark | Contributors | Deadline | Status | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
Train virtual sites and BCCs to existing HF/6-31G* data and additional Br data, with a weak constraint to initial charge | Compute additional multi-Br data | Nov 2023 |
| |||||||
Re-fit 2.1.0 BCCs to ESP data | Dec 2023 |
| ||||||||
Re-fit valence and vdW parameters to condensed phase properties | Re-fit FF terms for vsites-candidate | Mar 2024 |
| |||||||
Benchmark | Improved or equivalent performance for molecules with virtual sites added (Cl, Br, pyridines) on training data | Mar 2024 |
|