Issues/PRs | JW – Radicals PR - I think it’s about ready, wondering about safety valve like allow_undefined_stereo where we enforce a cordon only around loading and then stop complaining once loaded MT – That would clearly be a ton of work, and if it does end up being the case that some user needed us to support radicals, we can roll it back and cut a new release. But it seems unlikely and wehave a remedy ready, so let’s proceed until we get information that people need radical support.
JW – I saw your issue on the MiniDrugBank replacement. I was thinking that we could come up with a checklist of “chemical diversity” and then programatically find a set that has all of those features. MT – From another direction, we could spin up Dalke’s machinery to help with this. But the generated-based-on-defined-features would also work. Should focus on molecules of minimum size so AM1BCC tests don’t take forever. JW – If we have a checklist of chemical features, would we want each molecule to have only one feature, or might we need to challenge the sets by having mutliple? MT – I think there should mostly only be one-feature-per-molecule. And it would be good to have pretty good coverage of all the input features. JW – Could also have things that we know AREN’T valid - like XeF6 and radicals. Brainstorming feature list: File format: SDF Elements: CHONPS+halogens (F Cl Br I) + monoatomic ions from Sage + Mg2+ Stereochemistry: Everything from that stereo issue + pseudostereo (little r/s) Overridable failures: Undefined stereo (eg planar C(Cl)(F)(Br)(H) w/ no graph stereo, planar C(Cl)(F)(Br)(H) w/ atom-centered stereo in conn table, C(Cl)(F)(Br)(H) w bond stereo in conn table (keep thinking about this one)
Aromaticity: OE aro table – Rings Resonance structure: Long unsaturated chain/dye Polyatomic ions: [O-][H]? P([O-])([O-])([O-])([=O])? Bond orders: single, double, triple Formal charges: -2 to +2
|