2021-03-26 QCA Submission Meeting notes

Amide Torsion Set

Simon

• Lots of errors; not sure what’s going on

  • DD: I re-deployed PRP env; it’s a bit of a blunt instrument so many of these can be due to killing the old deployment

  • The tasks will re-run since they’ve error cycled

Tautomer enumeration

Josh

• fixed tautomer enumeration that was dropping the original molecule geometries in QCSubmit

• TG: for torsion indexer, needs specific indices, which is a problem for tautomer enumeration (atom counts change); so, I enumerate tautomers first, then do torsion assignment

New fragmentation method from fragmenter into QCSubmit

Josh

• JH: running on the JACs set to see what fragments we get via the Pfizer method

  • JH: fragmentation introduces some headaches on workflow component order; very order dependent, and not sure of a good solution yet

New dzvp basis with iodine

• New basis should be in next psi4 release; don’t have a conda package present for Linux yet

  • TG: not sure how we want to handle this; errored iodine cases will work if we re-deploy stack with new psi4, but completed (erroneous) iodine cases will be forever wrong

  • DD: first step, inventory datasets, assess which have at least one iodine-containing molecule, get counts from those of how many molecules; action depends on how pervasive problem is on our collection of sets

• TG: in the STANDARDS, we’ll want to make use of the flags we developed for this on the old datasets, such as non-compliant, wontfix, etc.

• DD: shouldn’t submit any iodine containing datasets until we have new psi4 deployed

Genentech set2

Pavan

• PB: please move this set to scientific review; no longer making progress; 50% completed, 50% errored

  • DD: please let us know if what completed is useful, as well as what you find on why there are so many errored cases

  • PB: there is iodine in the set; may be a factor

MM on CHO sandbox submitted

David

• DD: finally in on the server; Ben’s improvements to table indices gave enough performance boost for this to go through

Industry submissions

David

• Any reservations on submitting the 7 industry sets?

  • some contain funny elements like Si

• JH+TG+SB: filter out elements that don’t work with our FFs for a start; also really interested in having just one, citable dataset for the QM

• DD: will formulate a path given the objective: a single, clean, citeable dataset in the end, including all molecules from all partners. Ideally too, QM and MM included in the same dataset, even if QM is “pre-optimized”.

• SB: interested in getting rapid feedback with some of these datasets, so please do prioritize getting some of them

Database duplicates

Ben

• there are duplicate molecules (~200), cause duplicate optimizations, torsiondrives

  • punting on it for now; results from server not really being thread/process safe, column for hash not unique

Test server is up

Ben

• BP: test server is up, entirely disposable, feel free to throw small datasets on it for e.g. prototyping STANDARDSv3 implementation, client refactor