open force field

An open and collaborative approach to better force fields

Nov 19, 2019 Advisory Board Meeting

Posted on 19 Nov 2019 by Karmen Condic-Jurkic

The Open Force Field Consortium Advisory Board met on November 19, 2019. The minutes are summarized as follows:

Personnel update

  • Y. Qiu and D. Slochower have left their positions at UC Davis and UC San Diego, respectively
  • S. Boothroyd moved from the Chodera lab at MKSCC, New York to the Shirts lab at CU Boulder, Colorado
  • OpenFF is hiring more personnel, primarily software scientists

OFF toolkit update

  • Version 0.6.0 released, with support for LibraryCharges (SMARTS-based charge assignment)
    • Release notes here
    • Currently only TIP3P is available, open to contributions of residue SMARTS-based protein FF
    • Important to note changes in charge assignment:
      • All molecules must be assigned charges by some method (charge_from_molecules counts)
      • If partial charges don’t add up to formal charge, an error is raised
  • Upcoming features:

OpenEye dependencies

  • OE tools are used in some of the OpenFF workflows, but the long term goal is to make Open Force Fields and fitting infrastructure independent of OE licenses
  • The goal is to make Open Force fields usable without OE dependence as soon as possible and we are working hard to make that happen, while the infrastructure used to build force fields will keep the current OE tools incorporated for a while longer, depending on the available developer time
  • J. Wagner will build a converter between mol2 and sdf file formats (with charges) as a short-term solution for users without OE licenses

Protein-ligand benchmarking update

  • D. Mobley gave a quick update on the current state of protein-ligand benchmarking, where everything is more or less ready for free energy (FE) calculations, but compute resources have been somewhat scarce lately
  • Additionally, preparation of the systems and the state of FE workflows made this effort harder than initially anticipated. For example, using Amber TI required 1 month of human time to generate input files. The selected method was based on V. Gapsys’ non-equilibrium switching approach

Parsley feedback

  • Checking who has used Parsley and what would make it easier for pharma partners to try it – provided Jupyter notebooks were useful to play with Parsley and simple protein-ligand systems
  • A desired feature for some partners is geometry optimizer to allow comparison of energies for different conformers – this feature comes with high priority for toolkit development
  • The best way to get support if there are any issues with Parsley or the toolkit is to report it on Github repository and if any help is needed with in-house deployment of OpenFF tools, contact J. Wagner

Other topics

  • J. Rodriguez-Guerra is working on GBSA energy calculations support in OpenMm, which will hopefully become available by the end of the year
  • Eli Lilly, Cresset and OpenEye are joining the Consortium