Main objective:

Find and implement a fast and accurate method for assigning charges in OFF toolkit. Test the influence of charges and LJ parameters on force field accuracy / parameterization protocols. 

Project owner: @Michael Schauperl (Deactivated)

Objectives for 2020:

Team members: @Michael Gilson @John Chodera @Hyesu Jang



Current Projects

Current Projects


Project / Study




Problem Space

Problem Space

Scientific questions

  • Is the current charge model optimal?

    • Can we use ML for quick charge assignment?

  • BCC reparameterization

  • RESP:

    • High-quality bespoke charges via multiconformer RESP-like methods

    • How do alternate RESP models impact accuracy?

    • ESP data on QCArchive?

  • How much do virtual sites help?

  • How much does polarization help? Which model (Drude, point polarizable, fluc-q) is best, or how do they compare?

  • Do we have right LJ types / type refinement?

  • Fitting electrostatics to condensed phase data - how? Which parameters?

  • Benchmarking results for different charge models 

Infrastructure requirements

  • Library charge implementation in OFF toolkit

  • BCC support

  • RESP implementation / support

  • Graph convolutional network (GCN) charges and WBOs

  • Fast, free replacements for AM1-BCC and WBO

    • Conformer generation and quantum chemistry

    • Graph convolutional networks

Data resources



  • AMBER, GROMACS and OpenMM have capability of running simulations with off-site charges. The problem is that not all off-site charges will work in all packages. We got the specific ones we want implemented in OpenMM, but not all are necessarily available in AMBER and GROMACS..