2020-09-30 BespokeFit meeting notes

(General) – We’ll need to reschedule around Jamshed's new availabiltiy

JH – Met with pharma benchmarking folks

JH – Adding tests for bespokefit

JH – Working on benchmarking bespoke workflow

• JH – Some ANI scans of larger molecules take 30 minutes. Could be a convergence issue (our convergence criteria could be too strict)

  • DM – You should post a reproducing case on GitHub. They may already know about this.

DM – Sounds like Janssen stuff is going smoothly.

• JH – Took them like an hour or two to get set up.

• JW – Potential Galileo involvement in industry benchmarking?

  • DM – Even if Janssen decides against this, we should keep it as an option for other.

JH – Recently got support from Galileo for spinning up QCF server + compute, in a way that plugs right into bespoke workflow.

JH - Demo of bespokefit, mostly worked, some bugs

DM – In Parsley paper, we found one shortcoming was that we don’t include torsion error metrics. So make sure that your benchmarking includes that.

• JH – I should check that optimized geometries actually improve after bespoke fitting.

• DM – Yeah, it’s possible that we just shift the balance between “getting thr right optimized geometry” and “getting good relative energies”

DM – I will be absent at this time next week.

JW – Vacation plans?

• JH – A week around christmas, and I’ll be moving at some point in the next few months.

• JH – I’ll be transitioning to a fellowship under Danny Cole, so I’ll have a holiday gap at the end of November.

• JW – Curious to know how much you’ll be involved in openFF ecosystem and how much in a new ecosystem. Keep me in the loop.