Page Comparison

• DH – Working on torsiondrives. Comparing to OPLS3e, but Schrodinger is considering whether to let us. Trying to pull ligands out of PDBs, but having trouble. Getting access to a cluster with Vytas.

  • JW – Should we commit resources to developing PMX?

  • DH – I won’t be doing software engineering on it.

  • JW – Once the toolkit approaches feature-completeness

  • JW – PDB parsing issue –> Post to core devs or developers slack channel and we can look into it

• MT – Pushed a fix to qcengine, not sure when release will happen. System object work / SMIRNOFF data structures. Looking at different structures for System data PE calculation. Some of the useful matrix stuff will require energy evals from matrix representations. This will require building in a new MM calculator, since I don’t want to be playing ping-pong with OpenMM. I improved how Pint quantity arrays are stored, taking heavily from D Smith’s work in QCElemental. Some hairiness wrt to elements in numpy array vs. openmm quantitiys vs simtk quantitys.

  • JW – QCEngine release coming later this week. Let’s see if we can squeak this in.

  • SB – For energy evals, we could use Timemachine, since it’s apache licensed

• SB – Finished moving dataset curation stuff from nistdataselection to nonbonded. Improved logic for scheduling simulations and grouping measurements for simulation. Have an open implementation of AM1-BCC

  • Github link macro
    link https://github.com/SimonBoothroyd/am1bcc

• DD – Reviewed CIMH PR. Worked on QCArchive advancements and dataset curation (psi4 patch). Job management. Looking at ways to clear existing incompletes, and automate restarting.

• JW – 0.7.0 release finalization. Conf gen script / pharma partner giftgiving thoughts. Pfantander talk Wednesday.

• Known inconsistencies page?

  • SB – Could have a master page for known inconsistencies, and link to it in API docs whever behavior is possibly inconsistent.

  • MT – Lots of examples of gotchas in other software docs. We could look at these as examples.

• Gimlet preparedness?

  • We should talk with John and Yuanqing about repo ownership, and when to have the infrastructure team “take ownership” of it, since then it’ll be unambiguous who is responsible for validation and packaging.

Possible unit solutions:

• unit in string (val=4 * angstrom)

  • Most explicit and verbose, most expensive

• unit alongside string (val=4, unit = angstrom)

  • Balance of above and below

• unit in header (val =4, val_unit = angstrom exists somewhere in a subsection header)

  • Less expensive, less readable, less explicit

• implicit units, i.e. store everything as floats with units implied (assume angstrom at the interface)

  • Least expensive, least implicit, least clear to users