2023-01-23 Chemical Perception meeting notes

Open

Building a micromolecule FF from scratch

@Trevor Gokey

• Recording: https://drive.google.com/file/d/1jdfAMQRV-MTaJaQvUYEAADALlWdGGXMt/view?usp=share_link

• (slide 7) – PB – Same periodocity for sage and new FF?

  • TG – Short answer: no. I’ll talk more about this in a later slide.

• (slide 8 ) – TH – You’re using chemical space regularization, right? I see you’re usign “chemical objective”, which penalizes complexity of types, right?

  • TG – Right. The biggest contributor is te number of parametesrs, and also the complexity of the SMIRKS.

  • TH - So here where we see the chem obj jump to 16, is this due to the increase in the number of active periodicities?

  • TG – You may be on to something. But IIRC this particular case doesn’t correlate with the addition of periodicities. But relatedly: I count each periodicity as two params: for the period and phase - So they are heavily weighted in chemical objective.

• (slide 9) PB – whihc energy cutoff did you use?

  • TG – 40 kcal/mol (basically none at all)

• (slide 9) TH – Did you turn off attenuation completely?

  • TG + PB – No, it still attenuated on its way to 40 kcal/mol

• (slide 12) TH – I kinda agree with your assessment… (see recording around 33 mins for full discussion)

• (stopped taking notes around 35 mins, see recording)