open force field

An open and collaborative approach to better force fields

News
Science Updates

Jan 2019 Open Force Field Consortium Workshop

Posted on 6 Oct 2018 by John D. Chodera

Group photo

We hosted the first Open Force Field Consortium Workshop in La Jolla, CA on January 7-8, 2019. The meeting was open to Consortium members and collaborators, and parts of the meeting were recorded for posting online. In addition, slides discussing current Consortium research and future plans are given below.

Details at a glance

Agenda

Comments and questions for discussion during the workshop are invited in the Slack channels indicated after each topic; e.g, #project-planning for “Roadmap overview”.

Jan 7

Research topics: 45 min talk + 15 min discussion

TimeEventSpeakerSlack channel
7.45 - 8.00Introduction [PDF] [Video]Gilson
8.00 - 8.30Roadmap overview [DOI] [Video]Shirts#project-planning
8.30 - 9.30Short-term parameter optimization strategy: ForceBalance [DOI] [Video]Wang#forcebalance
9.30 - 10.00Coffee break#random
10.00 - 11.00Long-term parameter optimization strategy: Bayesian Fitting [DOI] [Video]Chodera#bayesian-inference
11.00 - 12.00Chemical perception and SMIRNOFF typing [DOI] [Video]Mobley/Bannan#smirnoff
12.00 - 1.00Torsion fitting and fragmentation [DOI] [Video]Chodera/Wang#torsions
1.00 - 2.00Lunch (catered)
2.00 - 2.30Group discussion to define ad hoc breakouts and breakout leaders/note-takers
Proposed Breakout Session 1 : typing via chemical perception, torsion fitting
Proposed Breakout Session 2 : useful physical property data, parameter fitting
2.30 - 3.00Breakout Session 1
3.00 - 3.30Coffee break#random
3.30 - 4.00Breakout Session 2
4.00 - 5.00Reconvene for breakout reports and all-hands discussions
Breakout session notes: Benchmarking Accuracy [PDF]
Breakout session notes: Property calculation framework [PDF]
Breakout session notes: Use cases [PDF]
5.00 - 5.15Roadmap recap and discussion (reviewing Slack input from remote attendees)#project-planning
6.00Group dinner in La Jolla at Rock Bottom Brewing

Jan 8

Research topics 45 min talk + 15 min discussion

TimeEventSpeakerSlack channel
8.00 - 9.00Charge models [DOI] [Video]Schauperl#charge-models
9.00 - 10.00Fitting non-torsion valence terms [DOI] [Video]Mobley/Wang#valence
10.00 - 10.30Coffee break#random
10.30 - 11.30Software architecture and infrastructure [DOI] [Video]Wagner#infrastructure
11.30 - 12.30Fitting condensed-phase properties [DOI] [Video]Chodera/Shirts#propertycalculator
12.30 - 1.30Lunch (catered)
1.30 - 2.00Group discussion to define ad hoc breakouts and breakout leaders/note-takers
Proposed Breakout Sessions: electrostatics, software preview / toolkit hands-on
2.00 - 4.00Breakout Session (with coffee)
4.00 - 5.00Reconvene and wrap-up discussion
Breakout session notes: neural network potentials [PDF]

Note: Information below this line is mainly relevant to planning and attending the workshop, which has already taken place.

Directions

The workshop will take place at the Jacobs Medical Center Community & Medical Education Center, Rooms 2 and 3, located at 9300 Campus Point Drive, La Jolla, CA 92037 (mailing address for the entire medical center), about two miles from nearby hotels.

From hotel, take e.g., Lyft or Uber to the main entrance of the hospital. Enter there and walk all the way down the long corridor, past the Cove cafeteria on your left, into the modern Jacobs pavilion. Continue walking past the elevators to the very end, and the meeting room will be on your right.
See photos of the route below.

Getting to the meeting rooms

Meals and breaks

Breakfast, lunch, snacks. and coffee will be provided at the meeting room both days.

The dinner on Day 1 will be held in a private room at the Rock Bottom Restaurant and Brewery, 8980 Villa La Jolla, La Jolla, CA 92037, two miles from the meeting location. Please use Lyft, Uber, or another carsharing or taxi service to get there.
We ask that each individual or group pay their own portion of the bill.

Other discussion topics for the Open Force Field Initiative science team

Potential software architecture and infrastructure discussion topics

  • Property estimation and automated parameterization (#propertycalculator)
  • Scaling up property estimation
  • If many dependencies are required, do we want to have multiple conda packages, or a single monolithic package?
  • What should the API for automated parameterization look like to interface with ForceBalance and Bayesian methods?

Other potential discussion topics

  • Which AMBER biopolymer force field(s) should we target in the first optimized release? The current recommendation in the AmberTools 18 manual is ff14SB for proteins, so unless there is significant objection, we should probably stick with this.
  • Which solvent model should we target in the first optimized release? TIP3P or TIP4P-Ew?