| |
|---|
Database generation | CD – I’ll contact SB now that he’s back from vacation, will ask him to enumerate protomers CD – I’m using BOTH QCElemental and RDKit to check for connectivity, seeing almost perfect agreement CD will add link to uploaded slideshow here Summary of big slideshow showing results for charges AAs/termini in the protein set: For molecules that DON’T experience proton migration: When we restrict the geometry (maxcyc=0), we get significantly worse charges (OE - AT av. magnitude of 0.05 or less to more like 0.10 or less) For molecules that DO experience proton migration: When we restrict the geometry (using maxcyc=0), the charge agreement between AT and OE generally gets a lot better.
CD – Is it right to just look at absolute deviation? Or is it more important to weight the error by the magnitude of the charge? JW – It kinda seems like it would be really important tog et the most charged atoms right. For things like dipoles, the most charged atoms will be the most important (or said another way, small errors on those atoms will propagate to the biggest deviation in net dipole, compared to the same magnitude of error in a less-charged atom).
OM – Could make a plot of molecules where there’s “Absolute deviation of original” vs. “absolute deviation of maxcyc 0” – Then we could bin it into “good in both”, “bad in one method”, or “bad in both” JW – I’m not sure how we should prioritize looking into QM-derived charges, since that could be a can of worms (there will be a large overlap between molecules that rearrange in AM1 and molecules that rearrange in more detailed QM)
|
Molecules with at lease one deviation of > 0.10 from OE value | original | maxcyc 0 | smirnoff 0 1 1 1 1 0 1 1 1 0 1 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 1 1 1 0 0 0 0 0 0 0 0 0 0 1 1 1 0 0 0 0 1 0 0 1 1 1 0 0 1 0 0 0 1 1 0 0 1 1 1 1 0 0 1 0 1 1 1 0 1 0 0 1 0 1 1 0 1 1 1 1 1 0 0 0 0 0 0 Final score: original: 12/31 maxcyc 0: 15/31 smirnoff: 15/31 |
To dos | JW will try to find code snippet to get final OE AM1 geometry, will send to CD if possible CD will try different ways to summarize data from the large dataset CD will ask SB to run new optimizations using protomers on Lilac, and ask for geometries back CD will start testing geomeTRIC methods, such as placing restraints on valence terms (like bond lengths, angles, dihedrals)
|