| |
|---|
General updates | New meeting time (1 hour earlier) as of today CD – My class rescheudle rearranged, could we do the odl meeting time? JW – I think CD’s previous results were compelling enough to recommend use of ELF1 in the toolkit. Opened a feature request including that here:
|
“Next steps” from last meeting | Why does the data range end at exactly -1.0? Is something getting filtered out by the workflow? Maybe try putting cyanate in (C#N-) CD will try running just AM1 optimizations to see which deviations persist (if the parallel offset lines disappear, it’s a BCC issue) CD will turn off symmetrization to see if the vertical lines disappear. For antechamber, could be antechamber’s -pl 0 or -eq 0 cli arguments (page 301 of http://ambermd.org/doc12/Amber21.pdf) CD – I concluded that OE and AM1 match perfectly if we turn off symmetrization and optimization.
JW will look into whether substituted indoles (indoles with extra Ns) are known cases of disagreement between aromaticity models/chemical representations. (Regarding ELF1 plots): Difficult to draw conclusions yet as much fewer molecules in this graph than in above
|
| CD – I made a database of ~50 conformers per molecule for X molecules. I used those 50 conformers as inputs to ELF1, using both OE and SB’s RDKit ELF1, and I looked to see whether they yielded the same charges. I looked for a correlation between conformer geometry and the partial charges it gave. I didn’t find any correlation. All of minidrugbank with 200 conformers Each conformer from one ensemble to the identically-indexed conformer from other ensemble
Do conformer RMSDs correlate with charges assigned to those conformers? JW – This may help us investigate “if the two conformer generators simply create completely different ensembles, will ELF1 still provide more consistent charges than randomly selected confs?” OM – ELF charges are least interacting. That should make them more likely to NOT be missed by a conformer generator. Blue points are random conformers from OE and RDKit compared to each other. Red is ELF conformers.
CD + OM – This correlation looks very weak. I don’t think that we should pursue this further. JW – I’d be interested to know whether the ELF1 conformer pairs have, on average, lower charge RMSEs than the randomly selected conformer pairs.
|
To dos for next time | JW – I’d be interested to know whether the ELF1 conformer pairs have, on average, lower charge RMSEs than the randomly selected conformer pairs. JW – CD, could you upload scripts for doing the connectivity checking, maxcyc=0 runs, and data gen/analysis? Even a gist would be great. And link to
In next meeting, we’ll agree on upload format for data and scripts for AM1 project. Any way to add data/check for additional significance in comparing ELF1 to random conformer selection. For example “does the number of conformers we feed to ELF1 affect the agreement between the methods?”. This will be important to understand as we move this toward production use.
|
Further projects? | |
