"all-to-all" ddE computation #171
Description
Thoughts on migrating some of the ddE analysis I experimented with in rc2 into yammbs? The workflow is below, happy to discuss:
- starting from the QM structure, minimize with FF (already handled by yammbs)
- compute "all-to-all RMSD", i.e. compute RMSD between all QM conformers and all MM-optimized conformers
- assign each MM-optimized conformer to nearest QM conformer, even if that was not the starting structure
- compute dQM energies from lowest QM conformer
- compute dMM energies from MM conformers, taking the conformer that matches the lowest QM conformer as the minimum. If there is no match, this set of conformers is excluded from the analysis. If multiple MM conformers minimized to the minimum QM conformer geometry, this set of conformers is excluded from the analysis.
- all conformers that minimized > 0.3 A away (or another user-provided threshold) are excluded from the analysis.
- ddE comparison proceeds from dQM and dMM as usual
IMO this is a fairly different analysis to the normal straight ddEs we do (this method understandably winds up dropping a fair chunk of the data) so perhaps could be an additional ddE method, instead of replacing the current one.
The scripts in the ash-sage-rc2 repo implement this in 2 stages, not super well documented I'm afraid.
1: the RMSD/TFD comparison: https://github.com/lilyminium/ash-sage-rc2/blob/main/04_benchmark/qm/get-all-to-all-rmsds-and-tfds.py
2. the ddE calculation: https://github.com/lilyminium/ash-sage-rc2/blob/main/04_benchmark/qm/get-all-to-all-dde.py