Participants
Discussion
JM: I am almost done with generating the datasets, hitting some troubles with matching the torsions I want to highlight, debugging the smarts torsion match now. Also, regarding the other work on expanding kval plots some molecules are failing in the datasets.
TG: One thing that comes to my mind is when you do chemical environment matches you have to add the colon and number tag on your smarts pattern, that might be the issue here, we can look into that.
...
SB: Chris raised this point before that fitting the LJ parameters won’t help but optimizing the 1-4 scale factors, which is out of scope for this work but definitely we should plan in the near future.
Action items
- Jessica Maat (Deactivated) will continue working on dataset generation. TG will help with any debugging.
- Jessica Maat (Deactivated) will touch base with Josh Horton on setting up bespoke runs.
- Pavan Behara will do an all valence parameters refit of simple_fit1.
- Pavan Behara will repeat the over-representation analysis for TFD and RMSD metrics.
- Pavan Behara will compare the residuals with different priors (2, 10) on fit7.
- Pavan Behara will compare the residuals on BM set of molecules that match TIGs.
- Simon Boothroyd will give pointers on analyzing the parameter gradients wrt molecule populations, and smarts pattern to filter out for intramolecular hydrogen bonds.
...