2020-10-13 All-hands meeting notes

Updates

• DM – Update on biopolymer hiring

  • DM – Making an offer to Chapin, he has competing offer, he’s available in March

• JW – Upcoming 0.8.0 release, N-1 CIMH

  • JW – Will be able to define one less chargeincrement than tagged atoms. This will enable naive optimization where chargeincrements remain valid.

• JW/TG – Vsites

  • TG – I think they work – followed all the specs, made some simple tests with energy calculations with OpenMM, checks out so far

  • JW – TG implemented a clever way to check for symmetries and equivalent cases when atoms have multiple vsites – all in documentation

• MT – WBO bonds will be in 0.8.0

• MT – Canary tests

  • JW – this could be used as a standard way to sanity-check FFs in the future

• DC/JW – Initial Amber FF release – all inconsistencies mostly in sidechains, giving similar helical propensities in simulations

  • JW – a ton of redundant parameters, need to find a way to recognize these patterns and that should cut down 1/2-2/3 of these parameters, or we could make changes to our machinery, there are several ways to go about it

• MT – Update on System? – slow on development side atm, trying to figure out specs, envision how our data should look like and then meet to see where the overlap is

  • Enabling conversion between the objects would already be a win, considering complexity of this topic

• DD – Show running datasets on qca-dataset-submission

  • Can see progress here:

• KCJ – Working on organization preparation, will send around some organization documents for review.

• KCJ – Will begin maternity leave in a month. Should be back in early/mid 2021.

• DM/PB – PB and JM are working on torsion WBOs. Will reach out for infrastructure support if needed.

Ions

https://openforcefieldgroup.slack.com/archives/C1907SGET/p1602184377011800?thread_ts=1602028825.004500&cid=C1907SGET

• DM – I had this topic on my checklist for a while and I wanted to move it from my to-do list to other people’s too. We’re planning to start working on water models in the foreseeable future and once we do, we need to start thinking about ions. At the moment, people are using ion models compatible with a specific water model. How should we parameterize ions so they don’t aggregate and behave well?
  
  

• DM – What do we think? How should we parameterize ions? MS will be thinking about it. DC, do you have thoughts?

  • DC – Nonaggregation is important but hard. It’s been hard to go after solubility. Ligand field molecular mechanics stress computational resources. If the metal center is doing anything other than being an isotropic sphere, it’s likely going to be VERY complicated to work with, andI don’t hink many people ahve dealt with it.

  • DM – I’ve seen some people adding a 4th order term to the 12-6 interactions for stuff like cation-pi.

  • DC – All these 12-6-4 potentials are still isotropic, it would be interesting to see how these things sample out in a long run, but it’s not an easy thing to do.

  • DM – What are Amber plans about ions? 12-6-4?

  • DC – PMEMD has a 12-6-4 potential, which can apply to a limited number of atoms in the system. It’s currently being used for ions.

  • DM – JC also mentioned virtualsites as a way to handle ions. Would that be a good starting point?

  • DC – Vsites may help. In case of monovalent ions, you need to differentiate it against water, which may not be the simplest thing to do. Ligand field molecular mechanics may be a good solution for this. Still a lot of arbitrary parameters in this kind of work. Other than getting solvation dynamics and (an)isotropic behaviour right, everything else will be a challenge.

  • KCJ – How can we capture coordination changes?

  • DC – Ligand field MM has some mechanism for this. Something about a 5x5 coordination matrix. There’s some numerical stability issues but work has been done to solve it.

  • DM – Interested in doing something that wouldn’t require dramatic changes to our software and other packages. I view ligand field method as something that would be on a longer term horizon, as it’s something that it couldn’t get used immediately. Right now, we need to make sure that we have a water model and a right dataset to build a consistent set of (monovalent) ion parameters. Eventually we’ll have to worry about coordinated metal ions. It should be easier for us to deal with this than Amber because we can introduce new bonds in an easier way. (Might be a good spinoff project for a PhD student).

  • TG – Do virtualsites on ions get treated as three-point models? Would vsites be based on arbitrary other particules attached to the ion?

  • DC – Physics (eg rotational momentum) could be messy.

  • DM – We can extend this discussion later, and folks can feel free to pursue this as a spinoff project.

• DC (from chat) – I think the real way to put virtual sites on an ion is to say that the ion is actually some spherically symmetric (ortahedron, cube) group of particles very near to one another (0.2A), then dress it up form there.